Agents moving between distant ranks in a SharedGrid

[ncollier]

"It seems that I am experiencing synchronization issues with agents. I believe that in certain cases, the agents may not be able to successfully change their ranks as they are moving through the shared grid. I would guess that the issue might be related to situations where the areas controlled by origin and destination ranks are far from each other."

Check that each rank, given a point, can determine the rank associated with that point. Perhaps can only determine adjacencent ranks using the MPI Cartesian topology.

[ncollier]

When agent moves, agent only becomes out of bounds if enters neighbors local_bounds. Need to update this check to check everyone -- start with neighbors and then check other ranks.

See distributed_space.h: 214+

When creating DistributedCartesianSpace -- get all ranks local bounds via MPI, after local bounds and neighbors is computed?

[ncollier]

Now checks all rank local bounds after checking neighbors.

• [x] Test with Sandro's example
• [x] Check to make sure that the data in out_of_bounds_agents is being decreffed appropriately.