numpy -> jax.numpy everywhere

[willGraham01]

Resolves #44 |

Switches numpy to jax.numpy everywhere in the codebase. We keep numpy.typing.ArrayLike as a typehint because it's still applicable to jax arrays anyway (and jax doesn't have a good typehinting library of it's own).

This was largely a like-for-like switch of np -> jnp everywhere, save for in test_parameter_node.py where I think we had both a bug and something incompatible with jnp.allclose.

• Bug: Previously we were doing allclose( node.sample(...)[0], [0.3]*10), however this only compares the first element of our generated samples to 10 values in the generated list. The reaplcement removes the 0-index fetch from the samples, effectively now doing allclose( node.sample(...), [0.3]*10).
• Incompatible: jnp.allclose can't handle when one of the containers is a list, so swapped out [0.3]*10 for jnp.full((10,), 0.3) which builds an identically shaped and filled array.

There's also the usual ruff formatting that it wants to perform now that imports have changed.