[CF-metadata] GIS versus CF conventions

[Jonathan Gregory]

To clarify, I think there are three separate issues:

* Aggregation over files, which means coordinates etc. might be missing in
the file containing the data. CF does not formally allow this at present.
Should it?

* Description of how grids are constructed when they are not lat-lon. This is
in addition to the lat-lon auxiliary coordinates and is the idea of grid
mapping (appendix F) but could be more general.

* Relationship between different grids, for the sake of software which wishes
to relate u and v or u and T, etc. Balaji's proposal is relevant to this.

Jonathan