Try to add a sweat test

[asinghvi17]

The idea is to try regridding each grid against each other grid we support to see where problems crop up. I think this found quite a few though

[asinghvi17]

Ok looks like we're going to want to:

• Perform quadrature over the example field in spherical barycentric space, to get the average value over the cell rather than the value at the midpoint (particularly applicable for the gulf stream field)
• Just use regular vectors of the right length to contain data and regrid into/out of, instead of using the native field types, since they are quite difficult to write to in practice
• Make sure all fields are shifted such that their minimum is 0, so we can test integral preservation more easily

These should let most of the tests pass, I hope...

[navidcy]

The CI passes?

[asinghvi17]

It does but the rtol is too high, 1e-2. I'd rather get the "right value" of the field analytically there.

[asinghvi17]

Yup definitely a lot of weirdness this is uncovering.

[asinghvi17]

Wait what?! It completely fails locally