OptimResult return type

[sebasv]

How do you feel about a default OptimResult return type for all minimizers? I was thinking about a struct that contains:

• information about termination reason,
• runtime,
• number of function evaluations,
• the gradient/hessian (if applicable),
• the minimizing point and
• the target function value

(this list is heavily inspired by scipy.optimize)
More options can be added as more optimizers are coded.

[to266]

I'm definitely not against. Currently it's not done just for simplicity reasons, but if users find that other outputs are also useful, it might be a nice thing.