[Mitgcm-support] checkpoint36

[mitgcm-support at dev.mitgcm.org]

checkpoint36 is done.

Concerning Borotropic Crank-Nickelson time stepping:
I had to apply the filter (FFT or Shapiro) to
uVel and vVel when the 2Dflow_Divergence is not 
fully implicit, otherwise, the barotropic adjustment is
is not consistent and the model blow-up (even with using 
a small time step that was stable without filtering).
May be we can think of a different implementation for this
part (filter + Crank-Nickelson).

Jean-Michel