use GSW gsw_ct_freezing to compute freezing temperature of sea water

[jornbr]

The benefit is that this works natively with conservative temperature and absolute salinity, just like the model itself (and presumably unlike the current simple linear approximation). The drawback is that this is likely more expensive, even if the more light-weight polynomial variant of the function is used. However, computational cost may be small still, as it is a 2D field.

[bolding]

Maybe a switch to call or not - for a tropical simulation it does not make much sense.
Then introduce a full 2D field similar to SST. And only done at every macro timestep.