Make offline 1850 compsets closer to when coupled to CAM by using CO2 from datasets rather than hardcoded values

In doing #3800 I noticed that when coupled to CAM we use CO2 from CAM, but for offline 1850 simulations we use a hardcoded value. In order to make these more likely to be synchronized we should change it so that we use CO2 datasets from DATM for 1850 (probably do the same for 2000 as well). Obviously for fixed dates these aren't likely to be very different. But, they are more likely to be updated when using datasets and we are trying to check that the datm datasets correspond reasonably closely to what CAM is using.

There is an inherent difference for CAM as they are using a dataset with 360 latitudes with monthly data, and we'll be using a single global value for CMIP7 that's yearly.