DM-44305: Implement Gaussian Process interpolation over bad pixels

[PFLeget]

See this ticket --> https://rubinobs.atlassian.net/browse/DM-44305

[cmsaunders]

This mostly looks good. I have a lot of comments, but they are mostly related to bringing this into conformity with the LSST stack coding standards, rather than algorithmic changes or anything like that.

The one major thing is that there need to be unit tests for InterpolateOverDefectGaussianProcess and the other new classes. The standalone functions don't necessarily need their own tests as long as they are run through the unit tests for the classes.

Here are a few other general points, which I included in the comments, but didn't make a note of every time I saw them:

• For some reason the github Actions are not complaining about this, but a lot of the docstring lines are too long. The rule from the Dev Guide is 79 characters: https://developer.lsst.io/python/style.html#docstring-and-comment-line-length-must-be-less-than-or-equal-to-79-columns. For lines of code, the maximum is 110 characters.
• In the docstrings, the variable type needs to be surrounded by backquotes, i.g. `int`.
• Also in the docstrings, the summary sentence should be in the imperative sense, and should be formatted with the opening triple quotes on the same line as the sentence.
• We try to be as consistent as possible with capitalization. You have almost everything in snake_case, but a few variables are camelCase. It would be best to change those to snake_case too.
• If possible, pass the logger into the class and use that instead if warnings. If not, I think it would be better to use the logging package rather than warnings, just because it is what is most used in the stack.