New Bayesian global optimizer

[sovrasov]

Hello. I have an implementation of one Bayesian-like global optimization method https://github.com/sovrasov/glob_search_nlp_solver. It can handle nonlinear objective functions and functional constraints which satisfy the Lipschitz condition. My implementation supports dimension < 6. According to the NLOpt documentation most of the other methods (like DIRECTl) doesn't handle non-linear constraints (supports only boundary ones). On unconstrained problems this method performs not worse than DIRECT (see paper)
I can prepare PR to NLOpt. Will it be useful?

[stevengj]

Sounds great!

[sovrasov]

@stevengj I'll try to submit a PR as soon as I study the internal structure of NLOpt.

[stevengj]

Closed by #194.