MPI 5.0 RC issues pp. 59-84

[wrwilliams]

Problem

• AtoU page 60-61 border: this reads as a deprecation without a deprecation #964
• Page 63 line 37: MPI_XXX_FLUSH_BUFFER and MPI_BUFFER_FLUSH, surely
• Page 68 line 34: is a single bullet-point algorithm for send really a "model implementation"? Do we provide a code example/reference implementation elsewhere? #965
• Page 69 line 22: should deadlock be indexed/hyperlinked to the relevant bit of 3.5?
• Page 81 line 19: why is persistent not italicized here?

[wrwilliams]

Notes on the struck-out entries:

• Semantic changes are out of scope for the RC. The AtoU is jarring because it goes right up to the line of "global buffers are deprecated, don't use them" but I assume we decided not to deprecate for good reasons.
• Making the pseudocode here a real "model implementation" is again out of scope for the RC. I considered and rejected renaming it to "model algorithm" (even less correct).
• The italic "deadlock" is correctly indexed and I guess we're using this as a main definition so no link.

[jeffhammond]

Globally-scope buffers are bad but were not bad enough to justify deprecating them immediately. We should do so at some point, probably in the next release.