The big board logic rewrite of 2024

[pkrisz99]

This is a chonky one.

Board logic changes:

• Finally using make-unmake
• No longer copying a giant hash vector every make move
• Encoding castling as KxR
• Actual "is legal" function for verifying moves
• Handle null moves separately
• (probably more, I don't remember)

The code is really bad currently, but at least it works. It will be cleaned up later.

Also includes some changes to bench:

• Uses more positions
• Resetting search afterwards
• Added longbench command

And some other unintended stuff that I uncovered while coding this.

Elo   | 5.36 +- 4.77 (95%)
SPRT  | 8.0+0.08s Threads=1 Hash=16MB
LLR   | 2.92 (-2.25, 2.89) [0.00, 5.00]
Games | N: 9148 W: 2132 L: 1991 D: 5025
Penta | [49, 1001, 2325, 1158, 41]
https://renegadedev.eu.pythonanywhere.com/test/65/

Bench: 2372395