[mlir][mpi] add support for MPI_Get_processor_name

The MLIR MPI dialect currently exposes operations corresponding to MPI runtime calls such as MPI_Comm_rank and MPI_Comm_size but does not currently provide a representation for MPI_Get_processor_name. It seems like this could be modeled similarly to existing MPI dialect operations and lowered through the MPI to LLVM conversion path.

I'd be happy to work on adding the operation and lowering support if this sounds reasonable.