[cr-xu]

An initial draft to add constraints for the mean optimizer in the pathwise/optimize.py.

This addresses the #14

Example outcome of constained vs. unconstrained optimization with an artifical exclusion zone.

[cr-xu]

A possible use case could be e.g. we want to find the best solenoid alignment with a reasonable QE on cathode.

Also, when we ran BAX experiments at APS with a safety constraint, the BAX steps were sampling points mostly satisfying the safety constraint (high charge), so the region violating it didn't really get explored. At the end with the get_bax_mean_prediction(), it chose the unxplored region due to a poorly-fitted mdoel there, which had no charge.
Adding this constraint also in the evaluation would make the final evaluation more meaningful.