Add optimization parameters to OptimiationStsate

[TorkelE]

Currently there are no way to access an OptimziationProblem's parameters from e.g. callbacks, as these are no part of the optimization state structure (this is generally the case throughout the SciML ecosystem).

One example where I would have had use of this: I would like to record the loss function progression across the simulation (ideally I think it would be nice to save this to the optimization solution through a keyword argument). And my plan was to store an empty vector as a parameter to the optimization problem and have the callback push the current loss value to it in each iteration. For single-start optimization I could have done this by just declaring any vector, however, for multi-start optimization I don't think this approach holds, as the callback needs to have access to something which is unique to the individual optimization runs.

[ChrisRackauckas]

Yes good suggestion. Definitely should be done. It's non-breaking to add more things here so it can be done any time. Good first issue.