Add Choclo as a required dependency of Harmonica and replace the current
kernels for prisms gravity with the ones in Choclo.

Allow the `prism_gravity` function to compute the easting and northing
components of the gravitational acceleration of prisms, using Choclo
kernels for it. Add tests that compare with bare Choclo results. Remove
the dispatcher and replace it with and `if`/`else` statement for
selecting the parallelized or serialized jitted function.

Add tests against dumb Choclo calls and a Laplace equation test for the
diagonal tensor components. Improve the warning on the docstring
explaining the direction of `z`.

The tests in Choclo already cover those and they are more extensive.

Check if the gravity tensor components of a prism are being computed at
any singular point. Choclo will return a ``np.nan`` on those points, but
it's worth warning the user about it before they get the results.
Add tests that check if the warnings are being raised on those cases.

Update the expected values after we updated the gravitational constant
in Choclo.

Modify the test function that compares prism forward model with the
analytic solution for an infinite slab: locate the observation point on
top of the slab.

Rename forward_func variable to avoid confusion with the forward_func
argument of the jit functions.

Move some details to a new Notes section, improve some definition of the
parameters, improve the warning regarding the upward direction, add
a References section.

Redesign the checker for singular points to make it simpler and fix the
issue with Numba that made some tests to fail.

Co-authored-by: Lu Li <[email protected]>

Run checks for singular points using Numba jitted functions to speed up
the for loops.

Add a section to the prism forward modelling page in the user guide
showing how we can compute all gravitational acceleration and tensor
components of a single prism.

…isms-choclo