NaNs in first time step of some nstep history fields

[mnlevy1981]

In the NUOPC cap (at least in CESM-MOM6), if I move tauuo and tauvo to a history file that is written every time step then the first time step will be entirely populated by NaNs. This is also the case for nutrient river flux forcing fields in the MARBL driver. For tauuo and tauvo, the calls to post_data() come from mech_forcing_diags() and in mom_ocean_model_nuopc.F90 that call comes immediately after the time step is incremented:

  OS%Time = Master_time + Ocean_coupling_time_step
  OS%nstep = OS%nstep + 1

For the river nutrient fluxes, I am reading in the data from call_tracer_set_forcing() (prior to incrementing the time step) and calling post_data() immediately... so I think there might be different mechanisms at play resulting in similar behavior.