Error checking MPI calls in mksurfdata_esmf

[ekluzek]

For issues:

#2744
#2502

in the fix we discussed adding a check for the return code from a MPI call. This would be best practice, but we also haven't consistently handled the return codes for MPI. So we should do the handling everyone in a consistent manner.

See this discussion:

slevis-lmwg#9 (comment)

[ekluzek]

Note also #2352 updating the MPI calls to the FORTRAN-2008 interface which likely handles errors in a different manner.

Also perhaps the shr_mpi_mod version of MPI calls should be done rather than the lower level MPI calls directly.