Add more Baltic test cases

[Hallberg-NOAA]

The Baltic test cases are good candidates for quickly and cheaply (the highest resolution uses 16 PEs, the coarsest just 4) testing multiple code variants in a realistic ice-ocean context. They are also very convenient for debugging, so if we detect code issues, this is my preferred test case for fixing them. We should expand our testing using these configurations.

The 4 existing Baltic test cases should be restructured as variants of a common test case (as is done for cases like flow-downslope and others, where there are z, layer, rho, and sigma test cases), although perhaps the 3 different resolutions might be kept separate since they use different input files. In particular, we should add a non-Boussinesq variant, along with other variants (e.g., using the various equations of state) designed increase the code coverage in our regression test suite. One other consideration with the new Baltic test cases is that they should be that they run for at least 2 days, because the first sea-ice only appears late in the first day.

[ashao]

This sounds like a great use of the Baltic test cases. I would suggest that when the test cases are consolidated that Baltic_OM4_025 and Baltic OM4_05 are maintained in a separate Baltic/OM4/ subdirectory. This would both simplify the input management structure and also might make it a reminder that those test cases are intended to mirror exactly the global OM4 configurations. For example, in figuring out the heat/salt budget diagnostics, I never had an instance where if it worked in Baltic_OM4_05 and not the corresponding global experiment. Furthermore it seems much more practical to perform the eventual diagnostic regression tests that we have previously discussed (and that I have a branch lying around) to be done in these small versions.