Reduce Regrid Work

[timothyas]

Closes #13

To summarize that issue: the regridding algorithm was regridding all data variables, then subsetting only the variables we wanted to store. This reduces any potential extra work, by first subsetting the variables we care about, and then regridding.

Also, I changed how the yaml file is configured. I moved "regrid" to be a subsection of e.g. MOM6Dataset or CICE6Dataset, since in my view this is really just one component of the workflow. I'm curious what everyone thinks of that. Otherwise, I think it's fairly straightforward.

[timothyas]

@danielabdi-noaa for some reason I'm unable to request you as a reviewer, but I'm wondering if you could review this whenever you get the chance. Thanks!

[danielabdi-noaa]

@timothyas Just tested this and it does what it is supposed to do. What I tested include:

• commenting out some of the "data_vars" entries, it does regridding only for active variables
• Warning printed when rotation file not provided
• Vector fields are included when rotation file provided
  Not really an issue, but when only "u0" is activated in "data_vars", "v0" is also included in the output due to the way
  the code is written.

I approve of this PR.

[timothyas]

Awesome, thank you @danielabdi-noaa! @frolovsa I'm requesting your review in case you want to/have time to take a look.

[timothyas]

I just chatted with Sergey and I'll go ahead and merge this PR.