Add capability for climatological pairing

[rschwant]

We plan to add capability for evaluation of climatologies, which would be monthly model average against a climatology of observations for that same month. This is a long-term goal. If you want to help add this capability into MELODIES MONET, let us know.

Some initial prototyping, see issue (#121) Should defer to post satellite merging,
as more design needs to happen to properly support all types of climatologies, e.g. seasonal, monthly, monthly-hourly (diurnal)...

This could also include capabilities like: Pairing model and observational data on a daily or monthly average time scale. We discussed that adding this in especially for the global models with monthly averaged output instead of hourly output would be particularly useful. This would need to be an option in the yaml file. This has been prototyped in the analysis_musica branch,
within scripts, run_analysis and supporting analysis_*py modules. Incorporation into the MELODIES-MONET driver will need
the time chunking loop.

Since this is a long-term goal, moving this to discussions, but if someone is willing to take this on we could move it to issues.

The driver is setup for direct pairing right now so this new capability would require driver restructuring.

[blychs]

Would the idea be to have a preprocessing capability? I am adding that for comparison with Pandora and with VOC canisters, and it is probably ready for a PR after v1 is realized.
The way I am doing it would certainly work for surface_obs and for products like the L3 MOPITT Monthly Means, but not for harder stuff like TROPOMI, which would need to use superobs or something like that.

The way my code works is:

• Adds a "preprocessing" keyword, right now with the options "calculate_columns" and "average". We could just lever on that for anything not requiring and averaging_kernel or regridding to specific scans.
• Pair the preprocessed data.

I don't see an easy, non per-case way of doing something like this with remote sensing products that require an averaging kernel, but any ideas are welcomed!

Cheers
Pablo