Add support for Jax.experimental.jet.

Is it possible to add support to diffrax for jax.experimental.jet? I would like be able to calculate higher order derivatives of the output of an ODE with respect to it's input, and repeated calls to the jacobian are obviously slow to perform. It seems like jax.jet provides a solution to this problem (see: https://github.com/jax-ml/jax/issues/5152#issuecomment-2636319226 and https://github.com/robochimps/vibrojet/blob/main/examples/jet_example.ipynb), as it is capable of reusing previous partial derivatives to calculate the next in a series. I understand that there is currently no jet rule for lax.scan and lax.while which might complicate matters, but even so is this something that could eventually be implemented into diffrax?

Start by reviewing diffrax's tracing and differentiation paths alongside the referenced JAX jet example and the lack of jet rules for lax.scan and lax.while. Determine whether higher-order derivatives through an ODE can be supported, then verify that jax.experimental.jet works for the relevant solver paths and repeated derivative calculations.