Tuning module

[N-Wouda]

See also #85. Most ALNS heuristics have a ton of (hyper)parameters. This happens more or less naturally, as may things can be tweaked - in operators, call-backs, but also in the core accept/stop/operator selection classes.

We can help users with tuning. The ML community has a lot of this already, with e.g. keras-tuner, ray.tune, etc. Those are used by a lot of people, apparently with some success. We could have a look at how these work, and take some of the good ideas for our own.

[N-Wouda]

Or should we just interface with one or more of those existing tuning packages?

[N-Wouda]

For now at least document how to tune a bit better, and provide some pointers to places that I know did this reasonably well.

[N-Wouda]

I think documenting this in #159 suffices.