Revision 214

fluxlat, not yfluxlat, should be set to 0 at the beginning of
clmain. So fluxlat is defined for a given type of surface even if
there is no point of this type at the current time step.

fluxlat is defined at each time step in physiq, no need for the save
attribute.