refactor(quant): M7.2a Composite Engine 模块化（quant/composite.py）

- cross_sectional_zscore / composite_score / build_factor_panels 从 local_engine 迁入
  quant/composite.py；新增统一入口 build_score_panel(daily, factor_specs)
- local_engine re-export 保持旧引用兼容；selection/engine 的评分面板构建均指向
  composite —— 选股与回测的复合分实现收敛于一处
- 回归：quant/eval/research/selection 一致性/qlib 引擎测试全过；全量 pytest 通过

backend/app/quant/selection.py

@@ -21,8 +21,8 @@ from app.domain.entities.selection import (
     SelectionResult,
     SelectionStatistics,
 )
+from app.quant.composite import build_score_panel
 from app.quant.factors import FactorError, compute_factor, get_factor
-from app.quant.local_engine import build_factor_panels, composite_score
 
 _UNIMPLEMENTED_DEFAULT = [
     "exclude_suspended 依赖停牌数据，当前未建模（结果可能包含停牌股）",
@@ -35,8 +35,7 @@ def score_panel_for_factors(daily: pd.DataFrame, factor_specs) -> pd.DataFrame:
     回测（LocalEngine）与选股（run_score_selection）共用同一构建 ——
     保证 v2 §25/§27「历史回测与当前选股使用同一套引擎」的一致性。
     """
-    panels = build_factor_panels(daily, factor_specs)  # 未知因子在此抛 FactorError
-    return composite_score(panels)
+    return build_score_panel(daily, factor_specs)  # 未知因子在此抛 FactorError
 
 
 def resolve_observation_date(daily: pd.DataFrame, as_of: date | None) -> pd.Timestamp | None: