Fix OPenMP issues in carchm

[jmaerz]

Hi @TomasTorsvik and @JorgSchwinger , in connection to PR #243, I checked the OMP private statement in carchm. Indeed a number of local real values were not included (I am aware that loop indices are private by default, but it doesn't harm to be even more explicit) - probably most importantly also for the Si units: rrho. The fix is only related to carchm. In combination with the fix in #243, variations in DIC inventories during the first time steps disappear. I hope, no further local real value escaped my investigation. Upon request, I sort in the values in the OMP statement (for now I thought it is easier to see the changes). Issues with the sinking in OCPROD still remain (I guess, the easiest would be to un-OMP the settling to avoid race conditions for k versus kdonor).

[TomasTorsvik]

Thanks @jmaerz !
it seems that some of these variables were missing in the private statement.
I believe the convention is that the index of the first DO statement inside a OMP PARALLEL DO is always private by defaault, but I can't find the reference now. Still, I don't think there is any harm in explicitly declaring them as private.