[MPI] Add MPI support to periodic triangulation

[eve-le-guillou]

Hi all,

This PR adds MPI support to the periodic triangulation.
This is in two fold:

• Implementation of a periodic ghost generator, to handle the periodicity at the boundaries.
• Implementation of all the MPI related methods of ImplicitTriangulation in PeriodicImplicitTriangulation.

The periodic ghost generator will add a layer of ghost vertices at the global boundaries, by copying data at the opposite side of the data set. The ghosts can then be treated like the ghosts generated by Paraview`s GhostCellsGenerator. The algorithm is implemented in the ttkPeriodicGhostsGeneration filter, however the filter will be called automatically when switching to a periodic grid using the ttkTriangulationManager. Therefore no additional step needs to be taken to execute this filter with MPI.

Both the meta grid and the local grid for the PeriodicImplicitTriangulation are PeriodicImplicitTriangulation objects. This is useful when the number of processes of the execution is low, and some opposite global boundaries are present on the same process. In this case, periodic ghosts are not created, and the local and meta grids are used to have the proper periodic behavior.

Some MPI methods in ImplicitTriangulation and PeriodicImplicitTriangulation were factorized in a new abstract class called RegularGridTriangulation.

Here is an example that creates a Wavelet dataset, makes it periodic and then smoothes the scalar field:

data = Wavelet(registrationName="wavelet1")

tTKTriangulationManager1 = TTKTriangulationManager(registrationName='TTKTriangulationManager1', Input=data)
tTKTriangulationManager1.PeriodicityinAllDimensions = 1
tTKTriangulationManager1.DataArrays = ['RTData', 'vtkGhostType']

tTKScalarFieldSmoother1 = TTKScalarFieldSmoother(registrationName='TTKScalarFieldSmoother1', Input=tTKTriangulationManager1)
tTKScalarFieldSmoother1.ScalarField = ['POINTS', 'RTData']
tTKScalarFieldSmoother1.IterationNumber = 5
tTKScalarFieldSmoother1.MaskField = ['POINTS', 'RTData']

[julien-tierny]

ok, thanks a lot @eve-le-guillou
the code for the periodic ghost did not look so straighforward in the end but you managed to make this work. congratulations!
I have a few remarks below. Thanks for addressing them

[julien-tierny]

hi @eve-le-guillou , thanks a lot for the fixes.

while reading again ttkIdentifiers.cpp, I realized that it only generated vertex identifiers.
I believe we should also generate cell identifiers (to make sure we can re-construct unambiguously the global identifiers of simplices of intermediate dimensions in explicit mode). could you add that up real quick please?
thanks!

[julien-tierny]

awesome, thanks. let's go!