albedo set to zero by start2archive+newstart

[abierjon]

Since revision r1944, start2archive reads the variable albedo (combination of albedodat=bare ground and albedos of CO2 and H2O surface ice) in startfi.nc, but doesn't output it in start_archive.nc
As newstart reads start_archive.nc and doesn't detect the variable albedo, this variable is thus not initialized, but still put in restartfi.nc file. For a lot of compilers, this non-initialization in newstart is corrected by initializing it at 0. Therefore, the restarfi.nc file after a start2archive->newstart contains a null albedo everywhere.

This is problematic for the first timestep only when running the model. Indeed, the albedo value is recomputed around the end of the timestep (once by albedocaps in co2condens, and later in physiq) and stored correctly for the next time step. Still, this produces a strong increase in tsurf at first time step during daytime (see joint figure).

[abierjon]

Fixed by r2630.