ICB: various minor questions/bugs (icbclv and icbstp)

[mathiot]

• In case of calving point over land point (ie, bad calving file), if no isf cavity, the model runs without any trouble. However, if the model runs with isf cavity on (with the same tmask(:,:,1), i.e. same coastline/ice shelf edge), the model stop because some icb are grounded in a land point.

To avoid this, I suggest to mask the calving variable with tmask(:,:,1) in icbclv.F90 like this:

      zfact = ( (1000._wp)**3 / ( NINT(rday) * nyear_len(1) ) ) * 850._wp
      berg_grid%calving(:,:) = src_calving(:,:) * tmask_i(:,:) * zfact * tmask(:,:,1)

      ! Heat in units of W/m2, and mask (just in case)
      berg_grid%calving_hflx(:,:) = src_calving_hflx(:,:) * tmask_i(:,:) * tmask(:,:,1)

This modification leads to the same behavior with and without ice shelf cavity (ie, mask the calving file and run without any sanity check).

• Is the hard coded 850._wp the same as the one defined in namelist_ref by the variable rn_rho_bergs?

• In icbstp.F90: the rst_write is based on nn_stock. Should it be based on nstock or lrst_oce instead?

As I am not an ICB expert, I let people comment on it.

These points are also affected the trunk.