Validation tests to demonstrate jax-cosmo

[EiffL]

I'm opening this issue to collect ideas for validation tests and demo.

Thanks to @chihway here are a few options that would progressively demonstrate the code:

• Step 1: A 2 params Fisher + MCMC, for instance on a simple cosmic shear 1 bin example. This builds confidence that everything is going OK
• Step 2: Make the problem more complicated with like sigma8 omegam, and just something that makes Fisher unstable, show that jax-cosmo is still fine compared to chains
• Step 3: A full 3x2pt beyond Fisher, maybe with VI, maybe with Bayes Fast

I guess all of this can be done with cosmosis, it's just a matter of defining the right setup... and checking that the code works against a non-CCL thing.

[EiffL]

I have moved that issue over to the paper repo: DifferentiableUniverseInitiative/jax-cosmo-paper#1

We now have everything we need for all of these steps :-)