refactor mpi_cluster

[wds15]

Summary:

The mpi_cluster resource in stan-math should be revisited and it's design possibly improved.

Description:

The suggested improvements which need consideration are

• better doc for listening_state method for why this returns by reference (or revise directly)
• consider Meyers singleton design to handle MPI resource better
• consider use of lock_guard instead of unique_lock (see for a discussion here: http://jakascorner.com/blog/2016/02/lock_guard-and-unique_lock.html)

Reproducible Steps:

NA

Current Output:

The current output. Knowing what is the current behavior is useful.

Expected Output:

Describe what you expect the output to be. Knowing the correct behavior is also very useful.

Additional Information:

Provide any additional information here.

Current Version:

develop

[SteveBronder]

@wds15 is this still an issue?