Optimize Model.fit to skip re-initialization when already fitted

[jgallowa07]

When Model.fit is called on an already-fitted model, we could optimize by:

1. Skipping re-initialization of jax_data_sets if the data has not changed
2. Initializing parameters from the previous fit results instead of starting fresh

This would improve performance when iteratively refining a fit or running multiple fit configurations.

Location: multidms/model.py:127-128

Related: This might be useful for ModelCollection workflows issue 175