[DOC] Document more env vars?

[leofang]

We should make a distinction between env vars used at build time and those used at runtime. Runtime env vars include:

• MPI4PY_FUTURES_MAX_WORKERS (now documented!)
• MPI4PY_FUTURES_SERVICE
• MPI4PY_FUTURES_SERVER_HOST
• MPI4PY_FUTURES_SERVER_BIND
• MPI4PY_FUTURES_SERVER_PORT

and build time env vars include:

• MPECC
• VTCC
• MPE_LOGFILE_PREFIX
• VT_FILE_PREFIX
• MPICFG (now documented!)
• MPICC (now documented!)
• I_MPI_ROOT
• MSMPI_INC
• MSMPI_LIB
• MSMPI_BIN
• CCP_HOME

Some of them are not mpi4py's env vars, but maybe worth mentioning?

[leofang]

Maybe we should also accept MPI4PY_CFG as an alias of MPICFG...

[dalcinl]

1. The envvars related to mpi4py.futures should be properly namespaced as MPI4PY_FUTURES_XXX, and the previous names should be deprecated. Let me look at this first before any attempt to further document them (see mpi4py.futures: Proper namespacing of environment variables #74). The really important one is MAX_WORKERS and its meaning and use is already documented in the mpi4py.futures section. The other variables are related to relatively obscure and not frequently used MPI dynamic process management I implemented and support just for completeness. I would rather prefer to remain silent for the time being.
2. About the build-time environment vars, I'm not really sure. As time passes, more and more users consume mpi4py from pre-built packages, specially on Windows, so I hardly see the need to document I_MPI_ROOT, and MSMPI_XXX/CPP_HOME variables. MPICC is already documented. I would say that MPE* (used to be available within MPICH) and VT* (used to be available within Open MPI) are legacy stuff at this point. That leaves us with MPICFG. This one is not documented at all. You are right, the name sucks. Again, its use is rather specialized. Anyway, we could discuss a better name (e.g. MPI4PY_MPI_CONFIG) and add a note in the install section.

[dalcinl]

As you can see, I always find good excuses to not write documentation.