partial cells and vvl

[jchanut]

Here are a couple of changes I have made to made both key_vvl and partial cells compatible.
The idea is still to perform a vertical interpolation of tracers at the last level as in the constant volume case. My feeling was to perform this interpolation not on horizontal surfaces but on z* surfaces (z*(i,j,k)=e3t_0(k)+sshn(i,j), this is my feeling so it is questionable). One still has to add the Jacobian correction (depending on the sla gradient) everywhere including at the bottom.
There are certainly other ways to do this.

I created a new routine in dynhpg.F90: hpg_zps_vvl dedicated to that special case (key_vvl, ln_zps=.true.). It should however work in the more general case where ln_zco or ln_zps=.true., with or without key_vvl.
One more thing: this only works with key_sigma_vvl and bugs on the vertical scale factors definition corrected in domvvl (see Gurvan's tickets).

Jerome

[jchanut]

Patch VVL-Partial cells

[jchanut]

The definition of z* surfaces I gave is wrong. This is rather:
z*(i,j,k)=gdept_0(k)*(1.+sshn(i,j)/H(i,j)), where H(i,j)=1./mut(i,j,:) the total depth with key_sigma_vvl.