Implement NSGA-III

[abeham]

NSGA-III is originally developed for problems with >3 objective, but is also an improvement over NSGA-II even for problems with 3 objectives. There exists a unified version of NSGA-III (U-NSGA-III) that can be applied to problems with any number of objectives which is probably interesting to implement in HL.

[gkronber]

r17231: created a branch

[jkarder]

Please accept this ticket and add a comment for every revision you create.

r17551: Create NSGA3 plugin.
r17557: Add parameters for algorithm.
r17558: Generate structured reference points and show them in Results tab.
r17559: Initialize and analyze solutions of first generation.
r17615: The list of Reference Points are now a storable field in the NSGA3 class.
r17616:

• Implement the first part of NSGA3.
• Solutions are saved in a List.
• Minor changes in Utility.cs.

r17617:

• The reference points and the fronts are no longer stored.
• Minor restructuring in NSGA3.

r17618:

• Add storable generation field.
• Add StorableType attribute to Solution class.

r17619: Implement second part of NSGA3. Add random field to ReferencePoint.
r17657: Implement recombination.
r17658: Add better error handling for MinArgMin and MaxArgMax methods. Add comments describing the methods.
r17661: Correct reference point generation.
r17662: Correction: Use NumberOfObjectives instead of Problem.Encoding.Length where it is correct.
r17663: Initialize population size with the correct number (according to NSGA-III paper).
r17664: Add generation count to results. Minor refactoring.
r17665: Refactoring: Move selection logic to its own static class.
r17666: Add Unit test project.
r17667: Bugfix: The correct number of reference points are returned on ReferencePoint.GetNumberOfGeneratedReferencePoints.
r17668: Bugfix.
r17669: Add scatter plot for resulting fitness values.

[gkronber]

@chaider: please test the implementation and check the code.