Different calculation for latent heat upper limit for ground water

[ccarouge]

In the latent heat code, we calculate an upper limit using:

CABLE/src/science/canopy/cbl_latent_heat.F90

Line 214 in 630e2e1

fupper_limit(j) = REAL(ssnow_wb(j)-ssnow_wbice(j)/frozen_limit)*frescale(j)

Because the ground water uses a different calculation for wb, we need to introduce a specific case for the gw_model modifying the upper line to:

     fupper_limit(j) = REAL(ssnow_wb(j)-ssnow_wbice(j)*den_rat/frozen_limit)*frescale(j)

Note: the difference in the calculation of wb and wbliq in the ground water is in question, so this issue should be paused while the question is investigated.

[ccarouge]

following the discussion in #558 , it is confirmed we will need to introduce a separate calculation for fupper_limit for GW and non-GW. It's probably best to introduce a ice_volume that is either wbice or wbice*den_rat and keep a single form of the fupper_limit calculation. This issue is then unblocked and can be implemented.

It is not sure if the frozen_limit should change in consequence to the change to the meaning of wb in GW. The only thing we can do is add a warning for the doc.

We also need to update the doc on the calculation of fupper_limit as the warning is outdated.