Adding tiebreaking to A*

[nluqo]

Found that small tweaks to A* can dramatically alter the performance in certain cases (linear visits in size of path down from squared). I wrote a few notes with demo here. I tried adding the least obtrusive of the changes (tiebreaking when f scores are equal) and attempted to write a test for it.

This is my first pull request for anything so I apologize if I'm missing something obvious.

[twpage]

Interesting! Excellent write up. It's been so long that I actually implemented my own A* - but I'm pretty surprised by these results.

[ondras]

Wow, nice and interesting! Thanks for the patch as well as the updated unit test.