pkg/shelfice: TKE BC at sloping top or bottom

[jm-c]

Currently, the wind-stress effect on TKE is accounted for by imposing a prescribed
value (function of uStar) of TKE at the surface, i.e., as a Dirichlet BC.

To account for the production of TKE due to friction at the bottom (i.e., bottom stress), the same BC could be easily implemented over a flat bottom.
However, over a sloping bottom, the C-grid current discretization (TKE @ wVel location) has a specific problem over a bathymetric "step" (kLowC(i-1) <> kLowC(i)):
For instance, assuming kLowC(i-1,i)=(7,9), in analogy with wind-stress, the X component of bottom stress at interface k=7 (function of uVel(i,k=6)) should contribute equally to production of TKE in column i-1 and i.
There is no issue for the first half to shallower column i-1 (can impose TKE value at BC location k=7)
but for the second half to deeper column i, the vertical location where a TKE Dirichlet BC should be applied is not clear:

1. imposing a value of TKE at the same level as it is produced (k=7) would fall "in the middle" of the water column where a full prognostic equation is solved for TKE.
2. imposing a value of TKE where the bottom is (i.e., @ k=9) would disconnect the production location (and dependence on uVel) from where it is applied.

None of these 2 approach seems good.

The same problem shows up under a sloping ice-shelf when trying to account for the production of TKE due to friction stress below the ice.