SST for FAFMIP experiments with conservative temperature incorrect

[russfiedler]

The SST passed to the coupler should be potential temperature. If using conservative temperature as the prognostic variable the default code handles this correctly. The FAFMIP modifications simply use the prognostic redistributed_heat variable as surface SST. Therefore a sanity check with zero perturbation won't reproduce unless the pottemp_equal_contemp variable is set to .true..

Either a new diagnostic variable needs to be carried or maybe we can get away with just a simple call at the appropriate time.