Optimize lens magnification Hankel transform

[bclevine]

Goal

The current implementation of the lens magnification Hankel transform in the _gamma_t_LSS function is slow. We should improve the performance by changing to an FFTLog based Hankel transform, such as the ifht provided by scipy.

Scope

Refactor src/dsf/data_vector/mag_bias.py to use FFTLog instead of direct integration.

This will be a $J_2$ transformation from $C_\ell$ to $\gamma_t$ ; we already have a $j_0$ transformation implemented in src/dsf/utils/converters.py as power_spectrum_to_correlation. We should therefore add another converter in this same style for the $J_2$ transformation which we can call in mag_bias.py. We can also consider moving both of the Hankel transforms into src/dsf/utils/integrators.py, and then adding wrappers in src/dsf/utils/converters.py. But this would probably be a separate issue + PR which we can address later.

Acceptance criteria

• _gamma_t_LSS is refactored and passes all tests/benchmarks.
• A new $J_2$ Hankel transform function exists in src/dsf/utils/converters.py.
• Documentation exists for the new $J_2$ Hankel transform function.

[nikosarcevic]

Thanks @bclevine

Can you also confirm if what we currently have in projections could be useful? We should think about how not to duplicate transform code.

If you can use existing transforms we can perhaps pull our prjections outside the covaraince into its own module? And if you cna improve the existing code to run faster that woudl be amazing! But of course if you need something specific then you could first add it to converters and then we can definitely move the proejctions out into its own module so we have src/dsf/proejctionsinstead of it being burried inside covaraince module.
WDYT=?

[bclevine]

In #30 I decided to first add new Hankel transforms in their own module without touching the covariance/projections code, just for the sake of getting the PR out faster. I'm going to work on the comparison with the covariance Hankel transforms as well but will make a new issue for that.

[nikosarcevic]

ok sounds good. small steps. you are of course correct.