Create evolutionary-algorithm-driven refinement method

[saxbophone]

• Candidate solutions would consist of vectors of the lengths of all the lines for a figure of a given shape.
• Fitness function should probably measure total area taken up by the resulting figure (less is better), but might want to consider the following alternatives:
  • Average length of line segments (shorter is better)
  • Longest line length (shorter is better)
  • Percentage of total bounding area utilised by the figure (larger utilisation is better)
  • Least empty cells within the bounding area (similar to previous metric but measures by absolute value instead of proportion)
• The following parameters should be configurable:
  • Mutation rate
  • Breeding population proportion
  • Number of Generations (unless a decision to use plateau-detection to decide when to terminate is used instead).
• Probably going to try and use Uniform Crossover at first, might change to k-point crossover (with K configurable) is uniform proves unsatisfactory.