Feature/point redesign

[Jammy2211]

In order to help @jhod0 begin implementing new point source function, I have done a redesign and refactor of the point module.

This does the following:

• Docstrings throughout the module, to make it clearer what different classes do.
• Incorporates magnification into the source-plane chi-squared calculation currently supported.
• Moves the image-plane chi-squared all-to-all calculation to the correct Fit object.
• Refactors FitPoint hierarchy to be clearer.

I am not overly keen on the design of the point/fit package yet, with inheritance used a lot and quite confusing. But I don't have the capacity for a full redesign yet, and think we should wait until everything works with JAX before moving more stuff around.

The merit of this design is that it mirrors other Fit objects throughout PyAutoLens, which is important for using functionality like the aggregator. It also means the API when you have a fit, e.g. after a model-fit, is quite clear and easy to use and contains everything:

fit = al.FitPointDataset(
    dataset=dataset,
    tracer=tracer,
    solver=solver,
    fit_positions_cls=al.FitPositionsImagePair,  # Different input to the one used above
)

print(
    "Minimum Distance Between Observed Multiple Images and Model Multiple Images Without Repeats:"
)
print(fit.positions.residual_map)

print("Log Likelihood Without Repeats:")
print(fit.positions.log_likelihood)

# IF FLUXES WERE INCLUDED

print(fit.fluxes.log_likelihood)

This should help us get started.

[rhayes777]

self.profile = profile or tracer.extract_profile(profile_name=name)

[rhayes777]

This extract_profile stuff looks a bit whacky to me

[rhayes777]

np.ndarray