Issues with Zonal, Meridional statistics preprocessors

[ledm]

This issue continues the discussion from #433 (now merged).

zonal_statistics has the following issues:

• It won't work for irregular grids, but that's not straightforward to fix - it currently raises an error.
• We may encounter grids that are regular in the latitude/longitude directions, but that vary with depth. It should fail if that is the case as it is not possible to perform this calculation at the moment.
• Many of the irregular 2D grids (like NEMO's ORCA1) are fairly regular for a lot of the latitude range. A possible improvement would be to add a y-direction option so that we could take the mean in the y direction instead of the latitudinal direction for irregular grids. This would work fine if you're only interested in a small zonal range (like the AMOC RAPID array at 26N, for instance!)

meridional_statistics has the issues:

• Similarly to zonal_statistics, it won't work for irregular grids - it currently raises an error.
• The mean 'operator' needs to weight according to fx data.

Also, these preprocessors are in the _area.py file. Is that the best place for them?