After test_set_interior pull request, create new MARBL issue

[mnlevy1981]

netcdf_comparison.py itself should be tested. A good unit test script would just create small xarray datasets with known pass / fail status in the comparison tool and make sure everything behaves as expected. Important edge cases to cover (in addition to the obvious checks):

1. datasets have no variables in common
2. NaN mismatch between two datasets
3. Datasets have a common dimension name but different values (e.g. both have dimension named num_levels but num_levels = 4 in one dataset and 5 in the other)

[mnlevy1981]

Another issue to create: move from init-twice to call-compute-twice for better test coverage (or maybe have both?)

[mnlevy1981]

Another issue: is it possible to clean up the error traceback code? Maybe goto statements or #include to cut back on the repetitive calls to log_error_trace() and return?

[mnlevy1981]

#13 and #18 should be addressed after the PR instead of during. Also, on the documentation front, "Write up POP's implementation of saved state".

[mnlevy1981]

Documentation clean up:

1. use new-found toggle ability (from regression_tests.rst) to hide longer output in init.rst
2. Better job documenting that we currently only compute interior tendency for one column at a time (might also need better comments in the marbl_call_compute_subroutines_drv.F90 documentation)