refactor: remediation plan Fase 0-4 completata

- Fase 0: env var allineate, SYSTEM_PROMPT deduplicato, .gitignore pulito
- Fase 1: GameContext isolato, ciclo DM unificato, provider thread-safe (contextvars)
- Fase 2: DB schema esteso (current_area_id, combat_state, character_items)
- Fase 3: thread-local connection pool + WAL + busy_timeout
- Fase 4: test suite (32 test, 100% pass)
- heuristic_fallback come parser di riserva (non secondo pilota)

src/engine/prompts.py

@@ -4,7 +4,7 @@ import json
 from typing import Any
 
 from src.engine import db as db_module
-from src.engine.state_machine import GameContext
+from src.engine.game_context import GameContext
 
 _KNOWN_TOOLS: dict[str, dict[str, str]] = {
     "roll_dice": {"die": "tipo dado (d3..d30, d%)", "count": "numero dadi (default 1)", "modifier": "modificatore (default 0)"},