Changeset 2037 for LMDZ5/trunk/libf/dyn3d_common

PVteta computed by PVtheta was not used. Also there were a couple of
problems with PVtheta:

-- PVtheta calls tetalevel in phylmd, and interpolates at

Earth-specific values of potential temperature.

-- PVtheta calls tourabs, and the computation of rot in tourabs should

be modified (it is not correct when there is a zoom).

-- Even when there is no zoom, the computation of rot in tourabs

should probably be modified: directly combine vcov and ucov, as in
tourpot, instead of dividing by cv and cu.