asymetric arrays lead to a crash. the asymetry was introduce
in the previous commit. This change fixes the crash and I think
it will keep the single cell case fixed too.

Making a vtkUnstructuredGridGhostCellsGenerator that doesn't rely on MPI
and using the object factory to make a vtkUnstructuredGridGhostCellsGenerator
that works with and without MPI.

Addresses paraview/paraview#18182.

Issue was with not keeping memory unmodified for unblocking sends
until send was checked to make sure it completed.

Addresses paraview/paraview#18175.

The fixed warnings were:

knownConditionTrueFalse,Filters/ParallelGeometry/vtkPUnstructuredGridGhostCellsGenerator.cxx:399,style,Condition 'inputGridCopy' is always false

useInitializationList,Common/Math/vtkQuaternionInterpolator.cxx:41,performance,Variable 'Q' is assigned in constructor body. Consider performing initialization in initialization list.

uninitMemberVar,Filters/Modeling/vtkContourLoopExtraction.cxx:210,warning,Member variable 'vtkContourLoopExtraction::DataSetBounds' is not initialized in the constructor.

uninitMemberVar,Filters/Modeling/vtkContourLoopExtraction.cxx:210,warning,Member variable 'vtkContourLoopExtraction::DataDimension' is not initialized in the constructor.

uninitMemberVar,Filters/Modeling/vtkContourLoopExtraction.cxx:210,warning,Member variable 'vtkContourLoopExtraction::Bounds' is not initialized in the constructor.

When some processes had ghost cells and cells and other processes
had no cells (and thus no ghost cells were needed or kept) the
processes with ghost cells would return early from the filter while
the other processes did not. This resulted in a hang. Also improved
the logic on when to compute ghost cells and when to use existing
ghost cells.

Cleaning up the documentation, removing unneeded options (since
the default is faster than the non-default cases) and improving
performance by sending less data, better memory management and
better cell adjacency queries.

Addresses paraview/paraview#17385.

The main improvement was getting rid of the AllGatherV for the
partition boundary. The test was also improved to check for
number of cells and bounds. This work is based on the paper:
M. Patchett, John & Nouanesengesy, Boonthanome & Pouderoux, Joachim & Ahrens,
James & Hagen, Hans. (2017). "Parallel Multi-Level Ghost Cell Generation
for Distributed Unstructured Grids" which was presented at LDAV 2017
(The 7th IEEE Symposium on Large Data Analysis and Visualization), At Phoenix, AZ, USA.

Addresses paraview/paraview#17385.

Adding in functionality to allow inserting vtkNew objects into
stl containers of vtkSmartPointer objects as well as comparisons
between vtkNew objects and raw pointers. Also removing Get() and
GetPointer() calls from vtkNew objects since in most instances we
can just pass in the vtkNew<> object instead of having to use the
Get() or GetPointer() methods to get the raw pointer.

This topic is the result of running clang-tidy to
modernize our usage of NULL and 0 to nullptr. It also
includes some manual and semi manual changes where
clang-tidy would not be expected to work (such as in
comments, or classes not compiled on this build)

There are definitely many comments and occurances that
this topic misses that we will need to fix over time.

This commit adds GetMeshMTime method to vtkUnstructuredGrid
and vtkPolyData.
It also switch some filters to ObjectFactoryMacro
for subsequent usage in ParaView

add casts in lots of places to mun

This commit reindents the code with the following utility:
Utilities/Maintenance/vtk-reindent-code.py

This utility changes the positions of the braces so that they are no
longer indented relative to the code block they reside in.

The bash command line used was the following:

for d in Charts Common Deprecated Domains Examples Filters GUISupport \
         Geovis IO Imaging Infovis Interaction Parallel Rendering \
         Testing Views Web Wrapping; do
  for e in cxx cxx.in txx txx.in hxx hxx.in h h.in c c.in; do
    find "${d}" -name "*.${e}" -exec \
      python Utilities/Maintenance/vtk_reindent_code.py {} +
  done
done

vtkPUnstructuredGridGhostCellsGenerator now supports explicitly asking
for ghost levels > 1. Previously, that was only possible when a
downstream filter asked for it.

SafeDownCast performs a series of virtual calls and string comparisons,
which is quite slow, especially when used in worker functions.

vtkArrayDownCast will switch between SafeDownCast and the more
efficient FastDownCast (only available for common vtkAbstractArray
subclasses) depending on whether or not FastDownCast is defined for
the specific array type.

Code extracted from:

    https://gitlab.kitware.com/third-party/proj.git

at commit 41bdf05361007f8c5186f3df9944d86bb273ed13 (for/vtk).

Updated vtkPUnstructuredGridGhostCellsGenerator to support an arbitrary
number of ghost cell layers.

- Do not compute ghost cells on an input data which already has ghost cells.
- Add a new BuildIfRequired option to specify if the ghost cells must be
computed only if required by the pipeline or even if it is not the case.
- Make it work even if some rank are feed with an empty input.
- Update the related test.

This filter is faster than vtkPUnstructuredGridGhostDataGenerator and
can generate ghost cells with or without global point ids.