Implement Static Exchange Evaluation (SEE) for Quiescence Pruning

[kylejrp]

Quiescence Search (introduced by #22) currently searches all captures without a tactical pre-check. This can lead to search explosions and wasted time investigating obviously losing captures.

Static Exchange Evaluation (SEE) gives a fast material-only tactical check to determine whether a capture is likely profitable. It helps prune losing captures before exploring them.

---

📌 Goals

• Add a SEE function for evaluating captures on a target square
• Integrate SEE to reduce quiescence search branching factor
• Prepare groundwork for future forwarding pruning optimizations

---

✔️ Acceptance Criteria

• Implement SEE(Board board, Move move):
  • Simulates attackers and defenders on target square
  • Orders attackers by least-valuable-attacker (LVA)
  • Tracks net material gain/loss as the exchange unfolds
  • Stops early when score is irrecoverable (negative beyond threshold)
• Integrate SEE into quiescence:
  • Skip capture if SEE(move) < 0
  • Still allow winning and equal captures
• Add debugging utility to log SEE values (optional build flag)
• Write targeted tests:
  • Positive SEE where recapture doesn’t happen (winning)
  • Negative SEE (losing capture)
  • Even SEE (trade is balanced)
• Run regression to ensure SEE reduces nodes without major Elo drops

---

🚀 Stretch Goals (optional)

• Use SEE in main search for:
  • Futility pruning near leaves
  • Bad capture pruning in move ordering
• Consider a margin (SEE < -50) instead of hard < 0
• Add SEE scores to move ordering heuristic

---

🔍 References

• https://www.chessprogramming.org/Static_Exchange_Evaluation
• https://www.chessprogramming.org/Delta_Pruning