Velocity oscillations above inversion

[pdziekan]

There are vertical oscillations of vertical velocity above inversion, e.g. in DryPBL or RICO.
This could probably be fixed by having perturbation of tht and rv as model variables instead of full tht and rv
(judging from the pbl libmpdata++ test, in which there are similar oscillations for boussinesq_expl, but not for boussinesq_impl).
To do this, it is necessary to add forcing proportional to the gradient of the environmental profile (the profile from which perturbation is considered).
Oscillations do not have large impact on flow statistic in the pbl test, so they are probably not a big issue.

[pdziekan]

NOTE from Wojtek: modeling perturbations instead of total values leads to problems with conservation of variables.