Add support for remapping to or from MPAS vertices

[xylar]

This is useful for visualizing and comparing quantities that live on vertices like vorticity and the barotropic streamfunction. It could also potentially be useful for cleaner visualization of coastlines.

ESMF will only create mapping files for corners (vertices) from meshes described in the ESMF unstructured mesh format. Therefore, this merge adds ESMF mesh output for all descriptor types. So far, it is only used for MPAS meshes but may be useful for other mesh types in the future. (I didn't initially realize you could mix and match -- ESMF mesh for the source but SCRIP for the destination.) The downside to the ESMF mesh format is that it doesn't preserve the regular structure of lon/lat or projection meshes, so it isn't very convenient for those and thus hasn't been adopted there.

[pep8speaks]

Hello @xylar! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2021-05-25 07:34:28 UTC

[xylar]

Testing

This approach was used to remap barotropic streamfunctions from MPAS vertices to a projection grid covering the Filchner-Ronne Ice Shelf. This approach worked fine for both remap and remap_file, and both bilinear and patch methods:

It also does no harm to existing MPAS-Analysis runs:
https://web.lcrc.anl.gov/public/e3sm/diagnostic_output/ac.xylar/analysis_testing/chrysalis/test_projection_dim_fix/