Initialize the physics grid directly from dyn_grid_init

[nusbaume]

Re-configured the physics-dynamics grid initialization for the Null dycore, so that the initialization is done by passing variables directly from dyn_grid_init (now renamed model_grid_init) to phys_grid_init, instead of using specialized subroutines and pointers.

Fixes #88
Fixes #90

Tests run:

Successful simulation of null dycore with Kessler physics suite on Cheyenne with Intel and on Izumi with NAG and PGI. The Intel and NAG test were also run with debug flags enabled. Finally, one simulation was done with debug print statements to ensure that the physics_column_type assignment operator was working correctly, and that the phys_grid_init routine was receiving the correct inputs.

[nusbaume]

@gold2718 Please let me know if these changes are what you generally had in mind. If so then I'll go ahead and add Cheryl as well. Thanks!

[gold2718]

Looks mostly good, just a couple of small fixes.

[nusbaume]

Adding @cacraigucar to review the code as well, if she wants.

Also, along with the requested changes I added some allocate checks as discussed in issue #91 . Let me know if this is not what you all had in mind.

[gold2718]

A few more changes.

[nusbaume]

I also made modifications for issue #90 . Given that it sounds like only to_lower and to_upper were being replaced, at least for the foreseeable future, I went ahead and marked the issue as being closed by this PR.

Of course if you would like to keep the issue open, even after these modifications, just let me know. Thanks!

[gold2718]

Suggested comment change and issue in phys_grid_init.

[gold2718]

Looks great now, thanks!