Make mesh files CF compliant

[xylar]

Meshes from MpasMeshConverter.x and MpasCellCuller.x failed the CF checker because they set Conventions = "MPAS", and no mesh creation tool wrote units on the mesh variables. This PR moves the metadata that Polaris currently patches in (E3SM-Project/polaris#784) into the tools that create the variables.

Changes

• mpas_tools.mesh.attrs (new): a table of long_name, units and standard_name for the MPAS mesh variables, adapted from Polaris's attrs.yaml, with latitude/longitude and cell_area standard names. It also adds cf_conventions() and add_mesh_attrs(). The table is a Python dict because pyyaml isn't a runtime dependency.
• MpasMeshConverter.x and MpasCellCuller.x (mesh_conversion_tools_netcdf_c):
  • They write Conventions = "CF-1.8 MPAS", keeping the entries (including any CF version) of the input's Conventions.
  • Every variable gets the attributes from the table. This fixes the wrong or unclear long names from Make mesh files CF compliant #760, e.g. cellsOnVertex was "vertices adj. to each vertex", the two triangle qualities shared one name, and the culler's dvEdge was "length of arc between centres".
  • ncutil::def_var() takes optional units and standard_name.
• make_planar_hex_mesh(), jigsaw_to_netcdf() and triangle_to_netcdf() add the same metadata and Conventions.
• SCRIP files: grid_imask units are "1" instead of "unitless".
• New tests in tests/test_cf_attrs.py check the Python tools and both C++ tools against the table. The docs are updated.

The legacy netcdf-cxx tools in mesh_tools/mesh_conversion_tools/ are unchanged.

Two commits only apply pre-commit formatting to modules that predated the hooks (planar_hex, triangle_to_netcdf and the SCRIP modules), so reviewing commit by commit is easiest.

Testing

I built the package locally with rattler-build (Python 3.14 variant). All recipe tests passed (70 passed, 5 skipped). I ran cfchecks (cfchecker 4.1.0) on the reproduction from #757 and on the QU 1920 km test mesh through the converter and the culler:

File	Before	After
planar hex base_mesh.nc	0 errors, 38 warnings	0 errors, 0 warnings
planar culled_mesh.nc (cull + convert)	1 error	0 errors, 0 warnings
QU 1920 km converted	—	0 errors, 0 warnings
QU 1920 km culled	—	0 errors, 0 warnings

While doing this I found a separate law-of-cosines typo in the converter's triangle angle quality, reported in #761 and not fixed here.

Fixes #757

---

Posted by Claude Code on @xylar's behalf. The testing, analysis and wording above are AI-authored; please check them accordingly.

🤖 Generated with Claude Code

[xylar]

@trhille, could I have you give this a look, since it will potentially affect MALI and Compass (as well as Polaris and Omega, which we are trying hard to make CF-compliant)?