OptimizationProblem type

[dpsanders]

It would be useful to have an OptimizationProblem type that encodes the function and original box, dimension etc.

[ericphanson]

Relatedly (and also kind of related to #44) I think it would be nice to have an OptimizationResults type that gets returned that would hold onto the problem data along with the results. What this would enable is you could call results = optimise(f, X) and then e.g. optimise(results, tol = ...) to tighten the tolerance without needing to repeat the work you already did; it can just keep subdividing from where it left off. This would be useful in conjunction with things like #44 so you can terminate after X number of function evaluations, check on the results, and decide to resume if you want to.

It could also hold statistics such as number of iterations, and perhaps a log of the min values (so you can see how fast it decreases; this could be enabled by keyword argument like it is in BlackBoxOptim). I think you could also support destructuring so that min_value, minimizers = optimize(f,X) still works as expected. And it could have a show method that formats the results nicely.

[dpsanders]

Sounds like a great idea!