Handling curvilinear grids for plotting

Some data output from ML (FastNetUK), when unflatted, is on a curvilinear grid (where latitude, longitude is a 2D array, and they both vary with latitude/longitude).

Options include regridding to a rectilinear grid as part of the read.py (slow).
Alternative is to be able to handle this through into plot.py. The latitude/longitude arrays need to be explicitely passed to pcolormesh, so we would need some logic to see if the grid is curvilinear.

If curvilinear, latitude and longitude dimensions are just monotonic integers 0... ndim, with 2D auxiliary latitude/longitude coordinates containing the rectilinear grid layout.

Start by reading read.py and plot.py to trace how latitude and longitude coordinates reach plotting, then inspect the pcolormesh call. Determine how the code identifies 2D auxiliary latitude/longitude coordinates and preserves existing rectilinear behavior. Done means curvilinear grids from unflattened FastNetUK output can be plotted with their explicit coordinates.

Written by the indexing model from the issue text.