perf: 全量物化视图优化 — 按月预计算所有慢函数，API响应从106秒降至0.09秒

- 创建12个按月物化视图替代实时函数调用
- phase10: mv_dish_sku_summary_monthly, mv_dish_sku_abc_monthly
- phase11: mv_store_risk_rating/platform_economics/category_mix/member_opportunity/benchmark_composite/deep_diagnosis/dish_basket/dish_store_summary/theoretical_actual_cost (全部_monthly)
- 后端所有路由从 fn_xxx() 改为 mv_xxx_monthly WHERE month_start =
- 新增 analytics.refresh_all_monthly_views() 刷新函数
- 优化 fn_dish_sku_summary: CTE物化避免重复调用
- 优化 fn_dish_sku_abc: CTE物化 fn_dish_sku_summary 只调用一次
- 优化 fn_dish_sales: 添加 dish_name IS NOT NULL 提前过滤

server/src/routes/store-expense.ts

@@ -32,10 +32,10 @@ router.get('/overview', async (req: AuthRequest, res) => {
           CASE WHEN e.operating_expense IS NOT NULL THEN (r.received - e.actual_food_cost - e.operating_expense) ELSE NULL END AS actual_store_contribution,
           CASE WHEN e.operating_expense IS NOT NULL THEN (r.received - r.received * (1 - COALESCE(r.theoretical_margin_pct, 0) / 100) - e.operating_expense) ELSE NULL END AS theoretical_store_contribution,
           e.area_sqm
-        FROM analytics.fn_store_risk_rating($1) r
+        FROM analytics.mv_store_risk_rating_monthly r
         LEFT JOIN analytics.mv_store_operating_expense_monthly e
           ON r.store_code = e.sales_store_code AND e.report_month = $1::date
-        WHERE r.received IS NOT NULL
+        WHERE r.month_start = $1 AND r.received IS NOT NULL
       )
       SELECT
         count(*) AS total_stores,