Communicator equality: same object vs. MPI_Comm_compare()

[drwells]

From #16369 (comment): I don't remember the details, but I ran into a circumstance in one of my projects in which I had to use MPI_Comm_compare() since there were duplicated communicators with the same network (i.e., either MPI_CONGRUENT or MPI_IDENT). We should probably assert equality through that function instead of just checking the pointers whenever we just need MPI_CONGRUENT.