Scaffolding for pluggable feature extractors

[drewoldag]

This PR touches a lot of files, but many of the changes will be something like "bazin" -> "Bazin". Because it touches a lot of files, I would recommend that this be a true "Pull request", and I would ask the reviewer to checkout this code branch and use it to run pre-existing notebooks or scripts to help confirm that everything is working.

I have added the tooling to make feature extractors pluggable - and I've updated all the places where I found FEATURE_EXTRACTOR_MAPPING dictionaries to use the dictionary that is now generated automatically when the various LightCurve subclasses are imported.

An open question that I'll turn into issue and address is a separate PR is: what are methods that subclasses of LightCurve must implement? It seems like most of the existing subclasses implement fit and fit_all. But it's not clear if there are more. We should add stubs to the base class and raise a NotImplemented exception for all the methods that a subclass must implement.

[github-actions[bot]]

Before [c683c9b0]	After [20226e50]	Ratio	Benchmark (Parameter)
failed	195M	n/a	benchmarks.peakmem_learn_loop('KNN')
failed	186M	n/a	benchmarks.peakmem_learn_loop('RandomForest')
failed	136±4ms	n/a	benchmarks.time_feature_creation
failed	149±2ms	n/a	benchmarks.time_learn_loop('KNN', 'RandomSampling')
failed	152±1ms	n/a	benchmarks.time_learn_loop('KNN', 'UncSampling')
failed	2.60±0s	n/a	benchmarks.time_learn_loop('RandomForest', 'RandomSampling')
failed	2.58±0.01s	n/a	benchmarks.time_learn_loop('RandomForest', 'UncSampling')

Click here to view all benchmarks.