[documentation] Add some information on running netpyne with MPI

[Ankur Sinha]

Please describe the documentation improvement you'd like to see.

Some documentation on using MPI would perhaps be useful, otherwise users have to go find other documentation to run netpyne using MPI. Maybe a couple of examples showing mpirun usage? Also, some information on how many cores to use for different network sizes. For example, if I have 40 multi-compartmental cells, should I use 40 cores or 80? Would using 80 make things faster? These decisions require one to know how netpyne creates the instance for MPI internally, but users shouldn't have to learn that.