BC for horizontal velocity at the surface

[yairchn]

Currently when computing the Shear term we use extrapolate as BC for the velocity ta the lowest level. It seems to me more physical that we should set the horizontal velocity at the surface to zero instead. Doing that results in behavior changes, see the buildkite for changes in the simulations in draft PR #1171.

In SCAMPY the shear is computed using the the ghostpoints here (similar in the TKE equation). Since U,V are symmetric they effectively have zero gradient at the actual surface (and non zero horizontal velocity).