Improved UODE inference of Lotka Volterra

[mattlevine22]

Improving #153 for UODE inference of Lotka-Volterra.

Made the following changes:

• Use Ensemble Kalman Filter (ContinuousTimeEnKF) instead of Extended, as it is more numerically robust
• Removed optimization scheduler---reverted to 4000 epochs of ADAM with Ir=1e-3.
• Fixed the plotter for comparing filtered means vs data---when using AutoDelta guide, svi_result.params doesn't always do what you think it does.

Additional thoughts:

• For fully-observed, low-noise, I would recommend the Discretizer + Simulator w/ DiracIdentityObservation pattern, as in https://github.com/BasisResearch/dynestyx/blob/1c02f609350a98d11f3212485b8ec9601bfbe02b/docs/deep_dives/l63_speedup_dirac_vs_enkf.ipynb

N.B. I took these settings (EnKF + AutoDelta with lr=1e-3) from the SINDY-notebook for Fitz-Hugh-Nagumo. It is reassuring that they generalized nicely to this case.

N.B.B. most of the 4000 epochs are to "hyper-refine" the inference; the data are explained quite well after 1000 epochs, and the final 3000 epochs mostly focus on full sparsification (to see this, rerun with 1000 epochs).

[DanWaxman]

I changed the base branch to #mh-uode and requested @markusheinonen to review -- that makes the most sense to me from a workflow perspective. Then afterwards we can work on merging mh-uode to main.

[DanWaxman]

These look good to me. In the spirit of faster workflow, I'll merge these into #153