Add new top level modular VTK cmakelist and module.cmake files and
restructure database to build in AVTAlgorithm classes into same library as
databases that subclass them.
Remove explicit wrapping and installation cmake commands since modular
VTK takes care of it.
Use VTK add library macros that old PV one was promoted too.
Change to use new export macro defined by the module.
Whitespace and ANNOYING_OLD_CMAKE_UPPERCASE_STYLE cleanup.

New structure is a into two modules. One is not wrapped and consists of
fundamental VisIt as a library functionality. The other is wrapped and consists of
the VTK level classes that are exposed in ParaView.

Change kit libaries to the appropriate VTK module.

vtkExportKit.cmake no longer exists in VTK so has been copied to
ParaView for compatibility to includes need to use this version.

Update uses of VTK_ADD_LIBRARY to call PVVTK_ADD_LIBRARY. PVVTK_ADD_LIBRARY
version is included in paraview for compatibility.

Linking AvtAlgorithms library to vtkParallel regardless of
whether VISIT_PARALLEL(i.e., MPI) is ON or not. This is needed
for using vtkAMRUtilities when constructing AMR datasets.

This allows us to accumulate VisIt's targets in order to export them.

Minor changes to remove an unneeded header include. Also fixed some declspecs on windows.

Converted to a single library so that it will be easier to create client server wrappings for ParaView.

Had to fix all the VisIt includes to work properly in static and as a shared library. Had to change the automatic reader wrapping to not use plugin calls. Fixed numerous other bugs that cropped up.

Currently time support has been aded to the parent algorithm so that we can support time everywhere. MTSD than just inherits from STSD and only has to redefine the requestdata since everything else is the same.

Added a parent class for all avt database formats which will contain
all the general functionality needed by children

The STMD algorithm now supports parallel reading of all mesh types. Updated the algorithm to remove unused points in each domain so that we reduce memory overhead on unstructured and polydata. Changed the Mutliblock layout so that each node is a multipiece dataset.

Currently we only convert files that have a single mesh to an AMR dataset. If we ever encounter a dataset with multiple AMR meshes we will have to revis this code.

Currently we only convert files that have a single mesh to an AMR dataset. If we ever encounter a dataset with multiple AMR meshes we will have to revis this code.