Rewriting annex - interoperability with other programming models [simplified version]

[minsii]

This PR simplifies the description of dynamic process and RMA sections.

• In the dynamic process section, we only make recommendation that user should avoid using it together with OpenSHMEM. This is because OpenSHMEM users tend to write SPMD programs and thus unlikely use dynamic process together with OpenSHMEM.
• In the RMA section, the semantics comparison is deleted. This is because most RMA program can be written using either OpenSHMEM or MPI RMA. There is no reason to use both models in the same program.

[minsii]

@jdinan In earlier thread WG meeting (04/02/2019 ?), you commented that (1) the interoperability description for dynamic process is redundant and (2) the memory model comparison for RMA might be problematic as there is no complete memory model defined in OpenSHMEM. You also suggested that we can simply say do not use two models together for these two aspects.

This PR simplifies the description following your suggestion. Could you please check if it works better for you ?

[minsii]

PDF for 2019-09-10 threads WG meeting:
main_spec.interop-simplified.pdf