Optimize backtracking and guesses

Closes #9

The approach outlined in #9 turned out to mostly result in significant performance improvements, which is not surprising. On rare occasions some puzzles end up taking significantly longer than they did using the previous approach, but this is likely due to situations where a particular incorrect guess early on leads you far down a path that results in nothing. That is the nature of this type of puzzle. In the majority of puzzles, this new approach is orders of magnitude faster. I tested with about a dozen puzzles of extremely high difficulty and 10/12 were much faster using this technique. Screenshots below of one of the puzzles showing the improvements:

Puzzle

Using old logic

Using new logic

You can see that this new logic typically results in significantly fewer calculations and much lower completion times.

mikemcbride / vue-sudoku