Score algorithm enhancement

[Doom4535]

Currently every time a child board is scored the Manhattan distance of each tile is computed; however, because we are only swapping one tile and the blank it is possible to limit this to just a score update. We could track the score of the parent and add or subtract the difference of the one swap to get the score of the child (this would scale exponentially, so for really large puzzles we could significantly reduce computation time).

[ermsdev]

I think this could be accomplished by somehow linking board states, which could also help prevent the weird illegal moves, but I'm not sure how to do that properly.