Replaced MPI_Init with MPI_Init_thread to avoid deadlocks from scotch

Issue/feature addressed

When a mesh in Hdf5 format was read in parallel, the IncNavierStokesSolver executable hang during the partitioning steps by Scotch.

Identified Backtrace:
SCOTCH_CALL(SCOTCH_dgraphPart, (&scGraph, nparts, &strat, &part[0]));
library/SpatialDomains/MeshPartitionPTScotch.cpp: 104
library/SpatialDomains/MeshPartition.cpp: 682

Test-case:
solvers/IncNavierStokesSolver/Tests/ChannelFlow_3D.xml

Environment:

• Hdf5 version 1.12.0
• scotch 6.0.6

Cause:
Scotch was configured with multi-threading support and the MPI communication was initialized using MPI_Init. However, inside the documentation by Scotch, it is recommended to initialize the MPI communication using MPI_Init_thread() instead of MPI_Init to avoid issues with threading.

Proposed solution

Change the initialization of the MPI communication following the guidelines in https://github.com/poulson/scotch/blob/master/INSTALL.txt

Implementation

Replaced MPI_Init(&narg, &arg) with MPI_Init_thread(&narg, &arg, MPI_THREAD_MULTIPLE, &thread_support) for the initialization of the MPI communication

Tests

Suggested reviewers

@dmoxey @ssherw @CFD-Xing

Notes

Similar issue reported in a different solver: https://develop.openfoam.com/Development/openfoam/-/issues/2791

Checklist

• Functions and classes, or changes to them, are documented.
• User guide/documentation is updated.
• Changelog is updated.
• Suitable tests added for new functionality.
• Contributed code is correctly formatted. (See the contributing guidelines).
• License added to any new files.
• No extraneous files have been added (e.g. compiler output or test data files).