feat: Adding new models module

[stevertaylor]

This is just a start. There's much that could be done here.

Example usage:
model_curn_vg = models.lhood_maker(psrs, wn_params, gamma_common=None, common_components=14, red_components=30, common_type='curn')

[AaronDJohnson]

It's a good start to something that will be very useful. I recommend moving to the ArrayLikelihood which uses vmap across pulsars with the same noise models to be much faster on a GPU for these models which meet the requirements to use it. I've recommended code which should work. There are a couple of design options that we could pick:

• psr.noisedict vs noisedict
• common_type string vs correlation function