refactor: 代码审查修复 - 日期过滤、性能优化、数据泄露防护

- 修复 data_loader.py 财务数据日期过滤，支持按范围加载
- 优化 MADClipper 使用窗口函数替代 join，提升性能
- 修复训练日期边界问题，添加1天间隔避免数据泄露
- 新增 .gitignore 规则忽略训练输出目录

src/data/storage.py

@@ -90,6 +90,113 @@ class Storage:
             CREATE INDEX IF NOT EXISTS idx_daily_date_code ON daily(trade_date, ts_code)
         """)
 
+        # Create financial_income table for income statement data
+        # 完整的利润表字段（94列全部）
+        self._connection.execute("""
+            CREATE TABLE IF NOT EXISTS financial_income (
+                ts_code VARCHAR(16) NOT NULL,
+                ann_date DATE,
+                f_ann_date DATE,
+                end_date DATE NOT NULL,
+                report_type INTEGER,
+                comp_type INTEGER,
+                end_type VARCHAR(10),
+                basic_eps DOUBLE,
+                diluted_eps DOUBLE,
+                total_revenue DOUBLE,
+                revenue DOUBLE,
+                int_income DOUBLE,
+                prem_earned DOUBLE,
+                comm_income DOUBLE,
+                n_commis_income DOUBLE,
+                n_oth_income DOUBLE,
+                n_oth_b_income DOUBLE,
+                prem_income DOUBLE,
+                out_prem DOUBLE,
+                une_prem_reser DOUBLE,
+                reins_income DOUBLE,
+                n_sec_tb_income DOUBLE,
+                n_sec_uw_income DOUBLE,
+                n_asset_mg_income DOUBLE,
+                oth_b_income DOUBLE,
+                fv_value_chg_gain DOUBLE,
+                invest_income DOUBLE,
+                ass_invest_income DOUBLE,
+                forex_gain DOUBLE,
+                total_cogs DOUBLE,
+                oper_cost DOUBLE,
+                int_exp DOUBLE,
+                comm_exp DOUBLE,
+                biz_tax_surchg DOUBLE,
+                sell_exp DOUBLE,
+                admin_exp DOUBLE,
+                fin_exp DOUBLE,
+                assets_impair_loss DOUBLE,
+                prem_refund DOUBLE,
+                compens_payout DOUBLE,
+                reser_insur_liab DOUBLE,
+                div_payt DOUBLE,
+                reins_exp DOUBLE,
+                oper_exp DOUBLE,
+                compens_payout_refu DOUBLE,
+                insur_reser_refu DOUBLE,
+                reins_cost_refund DOUBLE,
+                other_bus_cost DOUBLE,
+                operate_profit DOUBLE,
+                non_oper_income DOUBLE,
+                non_oper_exp DOUBLE,
+                nca_disploss DOUBLE,
+                total_profit DOUBLE,
+                income_tax DOUBLE,
+                n_income DOUBLE,
+                n_income_attr_p DOUBLE,
+                minority_gain DOUBLE,
+                oth_compr_income DOUBLE,
+                t_compr_income DOUBLE,
+                compr_inc_attr_p DOUBLE,
+                compr_inc_attr_m_s DOUBLE,
+                ebit DOUBLE,
+                ebitda DOUBLE,
+                insurance_exp DOUBLE,
+                undist_profit DOUBLE,
+                distable_profit DOUBLE,
+                rd_exp DOUBLE,
+                fin_exp_int_exp DOUBLE,
+                fin_exp_int_inc DOUBLE,
+                transfer_surplus_rese DOUBLE,
+                transfer_housing_imprest DOUBLE,
+                transfer_oth DOUBLE,
+                adj_lossgain DOUBLE,
+                withdra_legal_surplus DOUBLE,
+                withdra_legal_pubfund DOUBLE,
+                withdra_biz_devfund DOUBLE,
+                withdra_rese_fund DOUBLE,
+                withdra_oth_ersu DOUBLE,
+                workers_welfare DOUBLE,
+                distr_profit_shrhder DOUBLE,
+                prfshare_payable_dvd DOUBLE,
+                comshare_payable_dvd DOUBLE,
+                capit_comstock_div DOUBLE,
+                net_after_nr_lp_correct DOUBLE,
+                credit_impa_loss DOUBLE,
+                net_expo_hedging_benefits DOUBLE,
+                oth_impair_loss_assets DOUBLE,
+                total_opcost DOUBLE,
+                amodcost_fin_assets DOUBLE,
+                oth_income DOUBLE,
+                asset_disp_income DOUBLE,
+                continued_net_profit DOUBLE,
+                end_net_profit DOUBLE,
+                update_flag VARCHAR(1),
+                PRIMARY KEY (ts_code, end_date)
+            )
+        """)
+        
+        # Create index for financial_income
+        self._connection.execute("""
+            CREATE INDEX IF NOT EXISTS idx_financial_ann ON financial_income(ts_code, ann_date)
+        """)
+
     def save(self, name: str, data: pd.DataFrame, mode: str = "append") -> dict:
         """Save data to DuckDB.
 
@@ -104,13 +211,35 @@ class Storage:
         if data.empty:
             return {"status": "skipped", "rows": 0}
 
-        # Ensure date column is proper type
+        # 确保日期列是正确的类型 (YYYYMMDD -> date)
+        # trade_date: 日线数据日期
         if "trade_date" in data.columns:
             data = data.copy()
             data["trade_date"] = pd.to_datetime(
                 data["trade_date"], format="%Y%m%d"
             ).dt.date
 
+        # ann_date: 公告日期
+        if "ann_date" in data.columns:
+            data = data.copy()
+            data["ann_date"] = pd.to_datetime(
+                data["ann_date"], format="%Y%m%d", errors="coerce"
+            ).dt.date
+
+        # f_ann_date: 最终公告日期
+        if "f_ann_date" in data.columns:
+            data = data.copy()
+            data["f_ann_date"] = pd.to_datetime(
+                data["f_ann_date"], format="%Y%m%d", errors="coerce"
+            ).dt.date
+
+        # end_date: 报告期/期末日期
+        if "end_date" in data.columns:
+            data = data.copy()
+            data["end_date"] = pd.to_datetime(
+                data["end_date"], format="%Y%m%d", errors="coerce"
+            ).dt.date
+
         # Register DataFrame as temporary view
         self._connection.register("temp_data", data)