Fix bugs related to vertical lagrangian remapping

[cbegeman]

The critical change here is that remapping needs to be called at the beginning of split time integration in order to work in coupled model configurations, whereas before it occurred during Stage 3 of the split time integration schemes (after baroclinic and barotropic iterations).

The second important change is calling ocn_ALE_thickness from within ocn_vert_regrid to ensure consistency in vertical coordinate. layerThicknessTarget refers to the output of the vertical coordinate routine, which is used to compute the vertical velocity.

There was also a need to change loop limits to nCellsAll rather than nCellsHalo in mpas_ocn_thick_ale.F.

[cbegeman]

Testing

Internal wave test case with VLR active:

Differences are larger than machine precision, normalVelocity O(1e-5) m/s, layerThickness O(1e-2) m but behavior is similar:

Old

New