Gradient calculation and test

[Jacketless]

Functionality for field-level gradient calculation, including variable sample distances longitudinally by latitude, was required. This PR now allows users to call a Field object function returning new gradient fields in X and Y for the whole field, or a defined subset in space and/or time.

Added .gradient function to Field objects, using central
differencing and forward/backward differencing at edges.

Created test that creates small grid and compares internal Field
object gradient calculation to Numpy’s gradient function that cannot
use unequal sample distances.

[mlange05]

Ok, as a first step towards advanced field numerics this is really great. In the future we might want to extend this further to do on-the-fly gradient calculations within kernels, and we might also want to start thinking about how to generalise this for "all grid types", but this provides a really neat first-level test for further development. Thanks a lot @Jacketless ! IMHO this can go in as is.