Comms Optimization

[a-gilbert]

At the moment each MPI rank has a source, source buffer array, moment buffer, and potential array. Since blocking communications are currently used, we can eliminate buffers on all but rank 0--the program won't pass the communications (and therefore modify the potential, source, moments, etc.)

Further more, since only the potential or charge density is needed at any time, these two can be merged to decrease memory overhead.