vertical velocity and key_vvl

[jchanut]

In the variable colume case, the vertical velocity is updated for the next time step (n+1) by estimating the sea level at time step (n+2). This estimate uses emp flux at time step (n) and not (n+1). This is not consistent with the next computation of sea level.
One should move the call to wzv at the beginning of the time step, as soon as the new surface fluxes are available.
By the way, why do we need to save the vertical velocity in restart files with key_vvl ?
Jerome