Expandstartfi.e weird at poles

[jnaar]

When examining the field "area" with startfi files using the "expandstartfi" script, the poles are treated differently than everywhere else, with a larger area than their neighbors. Seems like something we wouldn't want to do since area normally varies with cos(lat).

[emillour]

The problem for the mesh area at the poles is simply that in the GCM physics this is a single (large) mesh. expandstartfi should account for that when splitting values over multiple longitudes.

Fixed with r2990