loading global meshes in parallel

Currently, loading a global mesh file issues an open call from each rank of the simulation to the same file.
This I/O pattern does not scale to large simulations.

Implement MPI-IO reading for the global mesh file read and collective reads, to ensure that reading global mesh files uses parallel IO

Locate the global mesh file read and the per-rank open calls, then trace how MPI is initialized around that path. Replace the global mesh read with MPI-IO and collective reads, and verify that all ranks load the correct mesh without opening the same file independently.

Written by the indexing model from the issue text.