Improve SolverTuner logic and unit tests

[AlessandroPierro]

Objective of pull request: Major update to the SolverTuner utility class.

Pull request checklist

Your PR fulfills the following requirements:

• [x] Tests are part of the PR (for bug fixes / features)
• [x] Docs reviewed and added / updated if needed (for bug fixes / features)
• [x] PR conforms to Coding Conventions
• [x] PR applys BSD 3-clause or LGPL2.1+ Licenses to all code files
• [x] Lint (pyb) passes locally
• [x] Build tests (pyb -E unit) or (python -m unittest) passes locally

Pull request type

Please check your PR type:

• [x] Feature

New Behavior

• The new API allows the user to specify a fitness function that evaluates the current set of hyper-parameters, based on final cost and number of steps to solution. This is the function that is maximized by the SolverTuner.
• The behavior of SolverTuner is now fully deterministic, thanks to the seed argument.
• SolverTuner now tracks all the hyper-parameters in a structured numpy array accessible with the SolverTuner.get_results() method.

Does this introduce a breaking change?

• [x] Yes

[phstratmann]

One additional point that came to my attention while Chinonso and I were tuning hyperparameters: In many cases, we don't want to search through a grid of hyperparameters. Instead, we would like to search through a set of sets of hyperparameters, i.e. we would like to try the combinations x = 1 and y = 2 x = 5 and y = 8 but not e.g. x=1 and y=8.

Could you add this functionality? Would be very helpful!