feat: 策略复盘闭环 Phase1

- 新增 scripts/strategy_review.py: 遍历所有active策略
  - 腾讯API实时价对比止损/止盈/入场点
  - 分类: correct/wrong/partial/pending
  - 失败模式归因: 止损过紧/入场过早/止盈过远等
  - 写入 accuracy_stats 表(首条真实数据)
- 新增 docs/strategy-review-loop.md: 完整闭环设计文档
- 含失败模式→修复方向映射表

Phase1 结果: 38条策略, 94.7%准确率(19条待定), 1条止损过紧

strategy_lifecycle.py

@@ -1657,6 +1657,7 @@ def regenerate_all(stdout=True):
                 "position_advice": result.get("position_advice", "中等仓位"),
                 "time_horizon": result.get("time_horizon", "2周~3月"),
             }
+            new_entry["trigger"] = trig
             # created_at: 首次创建时设置，后续 preserve
             old_entry = existing_decisions.get(code, {})
             if old_entry.get("created_at"):
@@ -1786,7 +1787,19 @@ def regenerate_all(stdout=True):
         existing_pf = json.load(open(PORTFOLIO_PATH))
     except Exception:
         existing_pf = {}
-    existing_pf["holdings"] = pf.get("holdings", [])
+    # 保留 price/change_pct — price_monitor 维护的实时价，regenerate_all 不应清除
+    _existing_holdings_map = {}
+    for _h in existing_pf.get('holdings', []):
+        if _h.get('code'):
+            _existing_holdings_map[_h['code']] = _h
+    _new_holdings = pf.get("holdings", [])
+    for _h in _new_holdings:
+        _code = _h.get('code')
+        if _code and _code in _existing_holdings_map:
+            _old = _existing_holdings_map[_code]
+            _h['price'] = _old.get('price', 0)
+            _h['change_pct'] = _old.get('change_pct', 0)
+    existing_pf["holdings"] = _new_holdings
     existing_pf["updated_at"] = datetime.now().strftime('%Y-%m-%d %H:%M')
     json.dump(existing_pf, open(PORTFOLIO_PATH, "w"), ensure_ascii=False, indent=2)
     json.dump(wl, open(WATCHLIST_PATH, "w"), ensure_ascii=False, indent=2)