Support wetting-and-drying with RK4 and subglacial thin film

[cbegeman]

This PR has two components:

1. Limit the land ice pressure applied to surface pressure in grounded ice regions
2. Ensure that grounded ice regions remain "dry" according to the wetting-and-drying factor.

For (1), landIcePressureApplied = min(landIcePressure, landIcePressureFloatation). landIcePressureFloatation = rho_sw * g * ssh_min where ssh_min = bottomDepth - h_crit and h_crit is the minimum thickness according to the wetting-and-drying algorithm. In plain language, in hydrostatic equilibrium, land ice pressure cannot cause the water column to thin below the minimum thickness.

rho_sw is the reference ocean density so our hydrostatic estimate for ssh_min could be quite off where the local ocean density deviates from rho_sw. For this reason, we also add the option to allow land ice pressure that exceeds this value by some fixed amount. I have seen some threshold behavior where simulations crash above a certain amount of excess pressure around 10m-equivalent ice thickness.

For (2), we estimate the thickness of the water column if we were to use the full ice pressure, applying a hydrostatic dSSH from the difference between landIcePressure and landIcePressureFloatation to the dynamic SSH. By incorporating the dynamic SSH we allow ungrounding only of cells that are already close to floatation.

There is a 3rd component which may be added to this PR but is not part of the minimum necessary changes to support RK4, wetting-and-drying, and a subglacial thin film. That is the update of the landIceFloatingMask/landIceFloatingFraction to allow melting in ungrounded regions. In the near term, we decided that would be cells where landIceDraft > -bottomDepth.

[xylar]

Thanks @cbegeman, this is really helpful!