Uncorrect REBUILD for restart file with only ocean subdomains

[lovato]

The restart file produced by the REBUILD TOOL do not account for the filling with some value of the neglected land subdomains, when the simulation is carried out only for the ocean subdomains (inimpp2).

In this case the rebuilt restart file presents NaNs? in all land areas, which are obviously not accounted in the loop over the single output files.

In turns out that using a single restart file produced by the rebuild to initialize NEMO with a different number of processors the set up of halos is not correct. This causes the failure of the model, e.g., in the schemes for computing the ssh.

A possible solution is to initialize to zero all the values in the ouput netcdf file, or a more refined one, which accounts for the domain decomposition in the rebuild tool and fill with zero only the land domain for which no output files are generated.