Methods like `smtk::model::Model::cellsAs<T>()` and
`smtk::model::Model::groupsAs<T>()` would append cursors
that were invalid rather than omitting values not matching
the container type. For example, calling `model->cellsAs<Edges>()`
should only return cells owned by the model that are edges
(as opposed to vertices, faces, or volumes).
Without this patch, invalid `Edge` cursors were being inserted
for each vertex, face, or volume in the model's set of cells.

When a model visibility is changed, if the model is a submodel of another model, we need to label the parent model to be modifed too. Also, when the visibility or color of a mesh being modified, the associated model needs to be labeled modified. This fixes ensures the client side of model manager still have property relationship of all entities and meshes.

Also removed a couple unused vars

When exporting and importing smtk models (.smtk) with meshes in json format, more logics are added so we
have control over whether native models will be saved or loaded when saving and loading smtk models. Also,
when saving smtk models with meshes, a mesh file (.h5m) will be written out for the meshes, and then
loaded back with the model when importing the saved smtk model.

Instead of having a SessionIOJSON class in every session for SessionIO support, the base
Session class now will handle export and import sessions, and writing and loading native
models by invoking "write" and "read" operators of the session. Also Made exodus SessionIOJSON
use new SessionIO methods

MSVC complains about it. It is used, so why GCC and Clang are happy is a
different story.

Older GCC apparently can't figure it out otherwise.

This one does the target name transformation.

Windows was getting undefined symbols  from the Python
wrappers without this.

On Windows, a boost::filesystem::path uses wchar_t which doesn't match
the type expected when using .c_str(). Instead, force it into a regular
string before use.

Windows is fairly strict here.

In order to make it easier to add future parameters to the view creation
mechanism, a new ViewInfo class has been defined.

Also added teh ability to specify the position of tabs in GroupViews

Was causing viewer to display abstract def instead of derived defs
This should resolve SMTK Issue #1

Previously when deserializing we didn't generate unique names, as we didnt
handle the use case of meshes already existing with names like "Mesh_1"
which would causing conflicts.

CMake does the rest.

When a Mesh Worker has finished it currently generates a temporary file
to store the mesh results. This file is now deleted by the MeshOperator, so
that we don't leave around bad mesh files.

In the future all this logic should be handled by remus, once it has
support for remote FileHandles, and remus::JobResult can list multiple files
generated by the worker.

Uses python bindings for attributes() and definitions() methods

The include interfaces need to use the build and source trees.

This was done so that it has the same behavior as the rest of the Import code.