Remove advice about local MPI_COMM_FREE

[wesbland]

The advice to implementors for MPI_COMM_FREE mentions this:

Though collective, it is anticipated that this operation will normally be implemented to be local, though a debugging version of an MPI library might choose to synchronize.

This text is confusing and somewhat meaningless as it basically removes the collective nature in advice. How the implementor chooses to have MPI_COMM_FREE is up to them as long as they implement the correct semantics.

[abouteiller]

I tend to think that this advice to implementors is somewhat useful in reminding that the spec does not command that COMM_FREE is a synchronization barrier (unlike say, WIN_FREE, which is).

[wesbland]

After more discussion, we decided to drop this issue.