Enable stochastic EOS with Bodner

[iangrooms]

This commit enables the Stanley parameterization for use with the Bodner23 MLE scheme. It also cleans up logic related to the Stanley parameterization throughout. Finally, it adds some documentation for the MOM_stoch_eos module.

[alperaltuntas]

@iangrooms

• Is this PR ready to be merged?
• This PR changes answers for NCAR configuration. Are the changes expected? And if so, could you summarize the reason of changes and whether there is a way to retrieve old answers (e.g., by setting runtime parameters to different values.)

[iangrooms]

Yes, this is ready. TL;DR: The answer changes are expected, and can old answers can be recovered by setting USE_STANLEY_ML = False.

The occasion for this PR is that we enabled and were using this stochastic correction to the equation of state (EOS) in the original, FFH MLE parameterization. When we switched over to using Bodner the EOS correction was no longer being used, as it wasn't implemented for Bodner. The Bodner scheme simply didn't use the EOS correction, regardless of how you set the input parameters. This PR adds the code to implement the EOS correction for Bodner. So our old results have USE_STANLEY_ML = True and aren't actually using the EOS correction, while with the PR and USE_STANLEY_ML = True it is using the correction and the answers change. Maybe it would help to note that with the old code if you set USE_STANLEY_ML = False it won't change the answers.

This PR also cleans up the logic and documentation throughout, but not in a way that changes answers.

[alperaltuntas]

I confirmed that USE_STANLEY_ML = False for all of our test cases. However, the restart reproducibility test fails with a runtime error at the initialization of the continuation run. Hopefully the currently open, relevant PRs will resolve that. Otherwise, we can revisit these changes. But for the time being, I'll merge this PR as is.

[iangrooms]

Sounds good. This particular stochastic parameterization has been passing exact restarts for a while, so I agree it's likely that the restart problem is related to the other stochastic parameterization (GM+E). If not I will revisit this.

[gustavo-marques]

Is the stochastic parameterization (GM+E) enabled in the tests? I did not think so...

[alperaltuntas]

It appears that ESCOMP/MOM_interface#267 broke the stochastic ERS test. This PR is good.