Incorporate freshwater fluxes into our forcings

[alxmrs]

From the monthly emulations discussion last Friday (as well as discussion with @adam-subel and @amogh-gulati at the m2lines offsite), there is consensus that it is essential to include freshwater fluxes in our forcing to improve the quality of our salinity predictions. In the status quo, it seems like our model may be attempting to predict salinity without knowing essential inputs (water coming from the atmosphere via rain), just from one ocean state to the next.

This issue has three parts:

• Track down (and likely regenerate) the data sources in our pipeline to see if freshwater fluxes are included in are training data. If they are note, we need to regenerate the datasets.
• Update our training process (likely: via a config) to incorporate freshwater fluxes in to our baseline training
• (Child issue): Run new training, eval, viz pipelines of our Samudra 2 / baseline model to see what effect this data has on the model (especially, salinity).