Adds Richardson number smoothing to CVMix

[vanroekel]

This PR implements smoothing of the Richardson number profile prior to computing
LMD94 shear instability mixing. This reduces the frequency of 2 dz (grid-scale) noise
in viscosity/diffusivity profiles

A flag for the number of smoothing loops is also implemented, but should not normally need to be changed from the default in registry.

Finally the limiting of the Ri profile to be positive is removed as this check is also performed in CVMix.

[vanroekel]

This has been tested in G and B cases within ACME. Both have run for 5 years. The zonal velocity cross-section along the equator is shown below for the un changed B (averaged over year 1)

with the changes in this PR, the result is (again averaged over year 1)

The observations are

Similar improvements are seen in temperature cross sections and meridional cross-sections of T and U.

[mark-petersen]

@vanroekel, thanks for your hard work on this. I tested with gnu and intel (debug) on wolf. Nightly regression suite is bfb on tests without cvmix, like the baroclinic channel. Global cases with cvmix on are not bfb, but differ by max 0.7 C in temperature between previous HEAD and this PR. I tried both 1 and 2 iterations. Looking at the code, I see there was no smoothing before, so there is no way to make a bfb match with the previous with any choice of the flag. That's fine.

[mark-petersen]

@vanroekel could you please delint? The easiest way is to open all the files you edited for this PR, and use the vi commands:

:%s/ *$//g
:%s/\t*$//g

to get rid of trailing white space and tabs. Then try (on IC)

wf-fe2.lanl.gov> alias linter
/turquoise/usr/projects/climate/mpeterse/repos/MPAS-Tools/MPAS-Tools/python_scripts/mpas_source_linter/mpas_source_linter.py
wf-fe2.lanl.gov> linter -f src/core_ocean/shared/mpas_ocn_vmix_cvmix.F

It will put the lint messages in the file src/core_ocean/shared/mpas_ocn_vmix_cvmix.F.e if there is any. I tried it and only saw two white spaces.

[mark-petersen]

@vanroekel good catch on that last commit. I agree that you should use all 'old' values when you smooth, not one new and two old.

[vanroekel]

@mark-petersen I think delint is done, let me know what else you need.

[mark-petersen]

Yes, thanks. Just trying to get people in the habit.

[mark-petersen]

@vanroekel and @jonbob: I am happy to merge this PR in as I think it has been sufficiently tested. However, because this is non-bfb, we need to consider when to merge it into MPAS, and if it should have a single commit into ACME.

[mark-petersen]

@vanroekel and @jonbob: Is this the final version you are happy with? If so, I will merge to ocean/develop locally, do a final test, and push it up.

[jonbob]

@mark-petersen - I believe so, but we'll wait for Luke to weigh in. It's what I used for my tests on anvil anyway

[vanroekel]

@jonbob and @mark-petersen , yes this is the final one I've been using in all testing.

[mark-petersen]

I tested one last time on the merge, everything looked fine. This also was tested in ACME for another 12 years starting at year 90 on an EC60to30v3 run on eos at:

/lustre/atlas1/cli115/scratch/mpetersen/a16q/run