Production Code Clean-Up

[ljlamarche]

This PR includes some major restructuring and reorganization of the code base, but should not add any new features or change how the inversion is done at all. This should make it easier for others to install and use this package, as well as organizing the modules to be more maintainable and flexible for adding features in the future.

Must be tested before merging.

[ljlamarche]

Comment from email with @almule12

• It looks like the script extends the inverted region beyond the FOV of the blue image, if I'm interpreting it right. This should definitely be fixed!
• Is the blue background subtraction implemented in your version?
• It seems like the geodetic output map for SigP is very large(256x256) even with decimation set high, do you know why that is?
• The one that immediately comes to mind is that your run script runs the gaussian denoising step for the blue imagery - I left that in as a comparison in my notebook but it serves no purpose in the run script, so I may as well take that out.

[ljlamarche]

@almule12 , regarding the FoV of the inversion, do we want it completely included within the blue image (i.e., no "empty" corners around the circular FoV)? Right now the limits of the common grid are chosen to be the limits of the smallest image, which is effectively the size of the original blue image in most cases.

[ljlamarche]

Regarding the blue background subtraction, I see that I'm calculating the backgrounds for all three colors (preparation.py lines 131-134), but I don't think I'm explicitly subtracting it anywhere. It's used various other places, looks like mostly as a filler for NaNs. Are you talking about just subtracting that value from the entire blue image early in the processing?

[ljlamarche]

At least on initial check, the background IS being removed in the to_rayleighs() function.

https://github.com/ljlamarche/asispectralinversion/blob/e7cd4cc21a27b5eb6c289f6a404ead307b242d79/src/asispectralinversion/preprocessing.py#L287-L304