Feature/xp no autofit import

[Jammy2211]

A recent PR on the child projects made JAX optional for likelihood functions, whereby users pass the JAX namespace as the variable xp through the source code.

This PR makes JAX optional at the highest level (e.g. PyAutoConf and PyAutoFit), including:

• For a non-linear search to use JAX, the use_jax input must be passed as True to the Analysis object.
• The non-linear search will internally work out if it supports JAX natively. This will ultimately have behavior where, for example, if gradients are used it uses jax.grad, if not it uses jax.jit, and if batching is support jax.vmap.
• Currently only Nautilus uses the Analysis.use_jax attribute to set up a jax.vmap.

There are few hacky unclean bits in the autofit model composition where it determines whether to use JAX based on input type. A more thorough consideration of how JAX should work in autofit will be performed in the future.