Oral Exam problem.

[martins-1992]

Currently the development is done on OralExamsTest#testCurrent.

• Implement this at OralExamsTest.
• Solve randomly generated instance.
• The problem should be solved via the ConstraintGroupBasedRepair.
  • Original version combines overlapping groups. Imitate that, in order to improve solution.
  • Implement custom constraint group optimization order. -> This is not needed.
  • Implement custom heuristics. -> This is not needed.
• Use Escalator optimization.
• Manually validate the solution. -> Check some parts randomly.
• Compare ConstraintGroupBasedRepair and hill climber performance.
• Make destinction between integration and functional test.

[martins-1992]

Currently, there is an issue, that an already used demand is tried to be allocated by the Repair-Optimizer.
-> That was caused by applying optimization steps in the wrong order.

[martins-1992]

Currently, the solutions contain only a couple of wrong allocations.
There seem currently to be 3 main differences to my original solution:

1. Prioritize plausible supplies for repair
2. Repair more than 1 group of the incorrect allocation.
3. Slight differences, how the optimization escalation was implemented.

I think the second point is the most useful one and easiest to implement.
The effect of the second one can be approximated via random repair with multiple tries, which is done.

I have already changed the group selector in such a way, that multiple groups can be selected at once.
Now I just need to implement a new group selector and check the results.

[martins-1992]

So I added the capability to repair multiple groups of a defiance at once,
but introduced the error of duplicate/too many optimization steps.
These need to be filtered.