Should Mesh class allow for any number of node coordinates?

[trexfeathers]

The UGRID conventions are flexible enough that mesh coordinates (node_coordinates, edge_coordinates, face_coordinates) can be pretty much anything - XY / XYZ / spherical coordinates / S2 indices / ... Any of these can be connected into edges or faces using the standard connectivity structure.

Iris' Mesh class currently assumes X and Y coordinates only. This enables a lot of assumptions during construction and when working with iris-esmf-regrid and GeoVista, but is much more limited than the scope of UGRID itself.

We could remove a lot of the structure of Mesh to allow for any sort of coordinates. If this makes downstream XY operations difficult then perhaps we could create a MeshXY subclass with all the extra XY conveniences?

This would be quite a big refactor and only something we could do before moving out of experimental (#5990) or in a future Major release.