The main bottleneck with real-time VTK rendering so far
appeared to be the normals computation. This commit addresses
that issue by making use of the latest normals computation
filter in VTK: vtkTriangleMeshPointNormals.

While this filter is much (5 to 16 times) faster than
vtkPolyDataNormals, it does not check for consistency in the
cell orientations that could cause inverted normals, which
is why the vtkPolyDataNormals is called once in the surfacemesh
renderdelegate to retrieve consistent cells for the input mesh.

See VTK merge request for more information :
vtk/vtk!2271

PS: That MR requires the latest commits from VTK master, which
does not include work made on texture wrap mode nor on multi
texture attributes yet:
- iMSTK/vtk@62a7ecd8
- iMSTK/vtk@ae373026

COMP: Fix return error in MeshIO::write

See merge request !123

System: gcc (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609

error: converting to ‘bool’ from ‘std::nullptr_t’ requires
direct-initialization

COMP: Fix function is not a member of std (Win)

The merge request !117 to add custom event handlers
for vtk interaction introduced a bug on Windows, where
the <functional> header needs to be included to be able
to use `std::function`.

This fixes the 50 errors listed on cdash (due to those
`std::function` members being used frequently):
http://my.cdash.org/viewBuildError.php?buildid=1111691

PS: The experimental build for this MR did not show any
issues on CDash [1], but they appeared once merged in the
master branch (continuous). There seem to be an issue
with how buildbot runs those builds on Windows, since a
similar issue happened on windows 32 bits with Eigen
aligned data structures in stl containers.

[1] http://my.cdash.org/buildSummary.php?buildid=1111615

See merge request !122

The merge request !117 to add custom event handlers
for vtk interaction introduced a bug on Windows, where
the <functional> header needs to be included to be able
to use `std::function`.

This fixes the 50 errors listed on cdash (due to those
`std::function` members being used frequently):
http://my.cdash.org/viewBuildError.php?buildid=1111691

PS: The experimental build for this MR did not show any
issues on CDash [1], but they appeared once merged in the
master branch (continuous). There seem to be an issue
with how buildbot runs those builds on Windows, since a
similar issue happened on windows 32 bits with Eigen
aligned data structures in stl containers.

[1] http://my.cdash.org/buildSummary.php?buildid=1111615

ENH: Add custom event handlers for vtk interaction

1. Allow to set up lambda functions from `VTKViewer`
as custom behavior to run in the interactor Style
in event callbacks.
2. Simplify `VTKInteractorStyle::OnChar()` to only have
simple default behaviors (start, stop, pause, play, quit)
3. Move the implementation of `getTargetFrameRate()`,
`setTargetFrameRate()` and `setSimulationManager()` from
`VTKInteractorStyle` to `VTKViewer`, which becomes friend class
of the private members of `VTKInteractorStyle` to avoid
repeatedly wrapping around the interactor style methods.

Co-authored-by: Hina Shah <hina.shah@kitware.com> (@hina), adapted from iMSTK/iMSTK!113

See merge request !117

Fix finite element deformation models and non-linear solvers

This MR does the following:

* Fixes internal force calculation for both semi-implicit and full implicit Newton iteration
* Fixes initialization of linear and StVK material models
* Adds a modified conjugate gradient solver to handle fixed boundary conditions
* Enables damping if provided in the config files
* Enforcing code guidelines and other clean up in the related files

These fixes concludes the refactoring and bug fixes for physics module. Once these changes are merged, one should be able to create any material model with choice of semi-implicit and full implicit Newton solvers while applying Dirichlet boundary conditions. In other words, all the vega functionality (barring reduced order models) can be reproduced in iMSTK.

See merge request !118

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

- default LinearSolver tolerance = 1.0e-4
- default IterativeLinearSolver maxIterations = 100
- remove those default values in ConjugateGradient constructor
as they are initialized in the parent constructors
- consistent use of this pointer

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

1. Implemented Homogeneous Dirichlet BC
2. changed `filter` lambda to normal method
3. simple cg solver implementation
4. return b.dot(b) instead of b.norm() within NewtonSolver::updateJacobian

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>
Written-by: Mohit Tyagi <mohit.tyagi@kitware.com>

Makes the update of LHS and RHS updateType dependent. For now only deltaVelocity type update is supported.
Clean up and renaming

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

1. Fixs the fully implicit nonlinear Newton solver
2. Completes implementation of Newton solver that can be either semi implicit (SI) or fully implicit (FI)
3. Adds options to set the solver to either SI or FI
4. Default is SI
5. Now the lamda function computing the internal forces needs to know if it is SI or FI
6. Prints warning if the FI Newton solver does not converge within max iterations set

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

Get rid of m_tmpStorage

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

- use m_vegaPhysicsMesh within initializeDampingMatrix
- consider all damping coefficients

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

- Invert parameters order in getInternalForce
- Only multiply by dT at the end

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

In FEMForceModel and StVKForceModel

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

Its name should be "CLFEM" and not "Corotational" to return
the correct type.

Written-by: Sreekanth Arikatla <sreekanth.arikatla@kitware.com>

Skip compute bounds in vtk

Apply changes referenced [here](vtk/vtk!2271) to speed up VTK rendering by avoiding calling `GetBounds()` which requires recomputing the bounding box at every frame for deformable objects:

1. RemoveCuller
2. Disable VBO Shift&Scale check

Also use `setUpMapper` (previously setActorMapper) more consistently along the vtk render delegates.

See merge request !115

Correct some warnings

Mostly due to lack of `override` where needed.

See merge request !116

1) Allow to set up lambda functions from VTKViewer
as custom behavior to run in the interactor Style
in event callbacks.
2) Simplify VTKInteractorStyle::OnChar() to only have
simple default behaviors (start, stop, pause, play, quit)
3) Move the implemenation of getTargetFrameRate(),
setTargetFrameRate() and setSimulationManager() from
VTKInteractorStyle to VTKViewer, which becomes friend class
of the private members of VTKInteractorStyle to avoid
repeatedly wrapping around the interactor style methods.

Co-authored-by: Hina Shah <hina.shah@kitware.com>

1) RemoveCuller
2) Disable VBO Shift&Scale check

Also use `setUpMapper` (previously setActorMapper) more
consistently.

MeshIO (input, output, convert) for (volumetric) meshes in  .msh and .veg file format

![liverScreenShot2](/uploads/f396f4cc1570700aa9c90e835dd932d2/liverScreenShot2.png)New addition(s):
1. Added `MSHMeshIO` to read a mesh in .msh file format, and instantiate `imstk::VolumetricMesh` 
2. Added a method in `VegaMeshIO` to convert `imstk::VolumetricMesh` to `vega::VolumetricMesh`
3. Added a method in `VegaMeshIO` to write `imstk::Mesh` in .veg file format.
4. Added an example `testMshandVegaIO` in `Sandbox` to test these implementations (Attached is the output with this merge request).

See merge request !112

Improve render clock

`VTKInteractorStyle::CreateRepeatingTimer` would call
`VTKInteractorStyle::OnTimer` then sleep for the period
given in ms after calling `OnTimer` again. This isn't
adapted for rendering at a uniform frame rate, because
the time spent within the `OnTimer` function is not
taken into account in the sleep time.

By using `OneShotTimer`, we can manipulate when to
render the next frame: in (period - time spent) ms
if the time spent in `OnTimer` is smaller than the period
time, or right away (period of 0 ms) if the time spent
is over the target period (late frame).

An interface is added in the `VTKInteractorStyle`,
wrapped in the `VTKViewer`, to set the target frame rate.

Also added a commit to use `DATA_DIR` defined in the Sandbox example to easily change the data directory for each user on their own machine. Any local change to the `DATA_DIR` value should not be committed in the future.

See merge request !114

vtkInteractor::CreateRepeatingTimer would call
vtkInteractorStyle::OnTimer then sleep for the period
given in ms after calling OnTimer again. This isn't
adapted for rendering at a regular framerate, because
the time taken withing the OnTimer function is not
taken into account in the sleep time.

By using OneShotTimer, we can manipulate when to
render the next frame: in (period - time spent) ms
if the time spent in OnTimer is smaller than the period
time, or right away (period of 0 ms) if the time spent
is over the target period (late frame).

An interface is added in the VTKInteractorStyle,
wrapped in the VTKViewer, to set the target framerate.

Default value is 0ms to render as fast as possible. See:
: http://gaming.stackexchange.com/questions/221423/is-there-any-reason-to-limit-my-fps

Current default to ".". Needs to be modified for each
user on their on system (do not commit changes to this line).

BUG: Fixes wrong default value for pbd tolerance

Fixes wrong default value for pbd tolerance in PbdConstraint class

See merge request !110

Fixes wrong default value for pbd tolerance in PbdConstraint class