Time-travelling ice

[kieranricardo]

For access coupling mode, this PR scales atm->ice fluxes by 1 / ice_fraction in med_phases_post_atm, prior to re-mapping atm quantities to the ice grid (time travelling ice approach).

For access-cm3 using parallel coupling, this reduces conservation errors by a factor of ~50, to approx ~0.01% (measured over one coupling period).

See also https://github.com/ACCESS-NRI/cm3-suite/pull/26 and https://github.com/ACCESS-NRI/cm3-um/pull/9.

[blimlim]

Thanks @kieranricardo for putting this together, it looks really good to me! Just had a couple tiny suggestions, but otherwise happy to approve

[blimlim]

I'm not 100% on this, but are the following needed for this subroutine:

   use med_map_mod           , only : med_map_field
   use med_internalstate_mod , only : mapconsf
   use med_constants_mod     , only : dbug_flag => med_constants_dbug_flag
   use med_methods_mod       , only : FB_GetFldPtr  => med_methods_FB_GetFldPtr

If they aren't needed, wondering if it's best to remove them just for clarity?

[blimlim]

Possibly unnecessary, but wondering if we should add the logging messages at the start and end just in case debugging is needed:

    if (dbug_flag > 20) then
       call ESMF_LogWrite(subname//' called', ESMF_LOGMSG_INFO)
    end if

...

  if (dbug_flag > 20) then
       call ESMF_LogWrite(subname//' done', ESMF_LOGMSG_INFO)
    end if