Theory prediction for miscentered haloes

[combet]

Add functionality to compute profile predictions for miscentered haloes, based on tests performed in the exploratory notebook explore_miscentering_theory.ipynb

• new functionality added to the theory module
• the theory demo notebooks (functional interface and OO) updated accordingly
• new tests implemented

[coveralls]

coverage: 100.0%. remained the same
when pulling 6ebe1aa on issue/588/miscentering_therory
into 721e2d9 on main.

[combet]

@m-aguena, @marina-ricci, @hsinfan1996 - following on the v2.0 release discussion and the idea of having miscentered profiles predictions part of the release, I'm thinking one compromise would be to implement the miscentered profile for a single halo (i..e not integrated over a miscentering distribution). This is in any case what we have currently in theory, i..e we don't deal with prediction for stacked profiles (that will most likely live in FireCrown/new CLcosmo prediction module?).

There are 2 routes to do so, both shown in the explore_miscentering_theory.ipynb

1. rely on the backends to compute Sigma and then the integrals brute-force (the very slow option I originally implemented can be sped up using @hsinfan1996's optimized integration for DeltaSigma without loss of precision)
2. implement the backend-independent version from @hsinfan1996, which uses analytical pre-computations of the miscentered surface density to speed up the process, i.e. backend-independent. This is the fastest, but needs more code to be added to the theory module.

I like option 2 better because of speed, but it would also make that part less consistent "in spirit" with the rest of the theory module that relies on the backends...

What do you think?

[m-aguena]

It all looks good. Thanks for all the hard work! (Just remember to clean the merge comment because it is huge)