Surfaces can compute more than they should

[unalmis]

We have things in the list of variables that we say we can compute on certain objects, but shouldn't be able. One could be misled that a quantity not typically associated with their surface was computed somehow by integrating on the boundary, or just by sampling the given surface.

If a quantity cannot be computed on an object, we should just not have a compute function for that object instead of setting dummy zero values.

Stetting dummy values causes bugs. For example n_rho defined as $\nabla \rho / \Vert \nabla \rho \Vert$ should not be defined on a ZernikeRZToroidalSection. But we say it is, and give it a dummy value of 0. Now, the area of that cross-section is well-defined, however we compute that area using n_rho, and since that dependency is a dummy for this surface, it shouldn't give the correct value.

[unalmis]

Remove @pytest.mark.xfail(reason="GitHub issue 1127.") from tests/test_configuration.py::TestGetSurfaces::test_get_zeta_surface once resolved.

[dpanici]

Fix for this is to remove the Surface level implementations of the compute quantities in question, and implement them only for the subclasses that actually make sense