Add co2_ppmv to atm2lnd_inst for consistent use in ctsm

@billsacks wrote in #1654 (I have paraphrased a bit):
When calculating co2_ppmv, pbot and pco2 should either be both downscaled or neither downscaled (currently pco2 isn't downscaled at all, but that needs to change: see #1332 ); either way I think you'll get the same result (within roundoff), because the downscaling should be done consistently for both pbot and pco2 (once #1332 is done). In fact, you might consider adding co2_ppmv to atm2lnd_inst and filling it in in lnd_import (in place of the current local variable in that subroutine) rather than recomputing the previously-held-but-discarded co2_ppmv value (and this would side-step the issue of whether to use the downscaled or non-downscaled version).

In #1654 I opted to use the not_downscaled version and to open this issue for a future refactor.