Warm nights wwc diag

[remicousin]

This is the first of the analyses where I needed to first write a function to then have groupby's map call it, as opposed to the other simpler cases. There is a bunch of others coming up. They will all look similar but I will make a PR for each to ease review.

The Definitions controls aren't pretty but I'll think about how to redesign it once all the controls are in there.

There is also the problem where the local plots draws 0s where there are no date (past 2014 for histo and pre 2014 for scenario). I had resolved that for the built-in groupby functions, but here I think I need to revisit my functions to see how they handle missing. I will do that now before submitting other wwc variables.

[remicousin]

Ok. The 2nd commit deals with NaNs vs 0s issue. I'll save for later the proper documentation and testing. That's it for this PR. In others using other functions, I might have to do similar changes.