Pass Sa_dens from atm to lnd for CESM

[billsacks]

As @olyson pointed out, Sa_dens is currently passed from atm to the mediator in CESM, but is only passed from the mediator to ice, not lnd. Instead, CTSM recomputes atmospheric density:

https://github.com/ESCOMP/CTSM/blob/4e126c75a2eaebc6c1c32da4f728bd0891599ff5/src/cpl/utils/lnd_import_export_utils.F90#L72-L73

This leads to some differences in the atmospheric density used in CTSM vs. the one sent from CAM – at least when running CTSM in a datm-forced CPLHIST mode (which is what @olyson was investigating). (We have not checked if the same discrepancy exists when running coupled to CAM. It could be that the discrepancy in CPLHIST mode arises from the fact that we're using 3-hour-average fields and so are calculating density from these 3-hour averages rather than instantaneous values; but it could also be that there are small differences in CTSM's calculation of density relative to CAM's.)

It seems like it would be best to ensure consistency between atmospheric density in atm and lnd by passing the atm values to lnd. I thought we might need to make some changes in CDEPS's datm to calculate atmospheric density for modes that don't provide it, but it looks like this is already done for the clmncep mode that is used to force land cases:

https://github.com/ESCOMP/CDEPS/blob/b5d4929ae2e6bdb8fa3f459ee3adc1a357e1b19f/datm/datm_datamode_clmncep_mod.F90#L467