Add Omega support to Cosine Bell tests

[xylar]

Checklist

• Testing comment in the PR documents testing used to verify the changes

[xylar]

Omega is using newer modules on Chrysalis than Polaris, so we need to update the Polaris version and build new spack packages before this can go in.

[xylar]

With E3SM-Project/Omega#169, this is working through forward steps but viz steps are failing because the dimensions are cannot yet be translated, see #172

[xylar]

Seems to be working!

Convergence isn't what it should be, so we need to look into that.

Viz is messed up but can be fixed (in collaboration with @andrewdnolan using https://github.com/E3SM-Project/mosaic/).

[xylar]

For comparison, MPAS-Ocean is performing considerably better, presumably because of higher order tracer advection schemes:

[xylar]

We should presumably do an MPAS-Ocean run with centered tracer adveciton for comparison.

[cbegeman]

We should presumably do an MPAS-Ocean run with centered tracer adveciton for comparison.

I'll do this. I originally thought we needed MPAS-O code changes but we might just need a few namelist changes.

[cbegeman]

I get the same convergence for MPAS-Ocean (tracer adv order 2, monotonic off):

[cbegeman]

I think this looks great! I ran it for MPAS-O; I think your Omega testing is likely sufficient. Thanks, @xylar

[cbegeman]

I will plan to add a separate test for 2nd order advection in a follow-on PR for benchmarking Omega against MPAS-O.

[xylar]

Okay, this runs fine (still fails analysis step) after a rebase.

[xylar]

Thanks again @cbegeman!!