refactor: internal board rep

[dannyhammer]

resolves #70

This PR migrates the contents of the chessie crate into the engine directly. A new module, src/board/ has been created, and all board representation / move generation / core functionality has been moved there.

In addition, the Game struct now stores its evaluation, which is updated on calls to make_move. So, evaluation is now computed incrementally, rather than being recomputed on each call to Evaluator::eval. Speaking of Evaluator, it no longer exists.

Not a huge gain, which is disappointing, but Elo is Elo.

---

SPRT:

Elo   | 8.81 +- 6.97 (95%)
SPRT  | 8.0+0.08s Threads=1 Hash=16MB
LLR   | 3.04 (-2.94, 2.94) [0.00, 10.00]
Games | N: 5994 W: 2419 L: 2267 D: 1308
Penta | [332, 433, 1358, 499, 375]

https://pyronomy.pythonanywhere.com/test/132/

bench: 11371856