refactor(ingest): unique key (site,handover_no,waybill_no) + business_date purge + shortfall-sum undelivered

- schema: expected_record unique key (site,waybill_no) -> (site,handover_no,waybill_no),
  so the same waybill can belong to multiple handover batches without being overwritten
- store: dedup by (handover_no,waybill_no) instead of waybill_no alone; UPSERT conflict
  key matches the new unique key; handover_no no longer force-overwritten on conflict
- store: add purge_expected, called on force re-download; deletes by business_date
  (download day, same source as task date) rather than batch_out_date, which is offset
  from the download day for sites like yunda where download day != outbound day
- store/runtime: thread force + target_date through ingest_task / _persist_to_db /
  _site_undelivered_handler so force re-download purges then re-ingests in one tx
- db_compare: undelivered_pieces now sums per-waybill shortfall (total - arrived)
  instead of expected_total - arrived_total, unaffected by unrelated/extra scans

inbound_verify/runtime.py

@@ -506,7 +506,7 @@ def _site_undelivered_handler(site):
 
             if store.ingest_enabled():
                 store.ingest_task(
-                    site, "undelivered"
+                    site, "undelivered", force=force, target_date=date
                 )  # 4 站 = ingest expected + actual
                 print(f">> [入库] {site} 前置入库完成")
         except Exception as e:
@@ -693,10 +693,11 @@ def _refresh_ready(site):
     _apply_ready(site, flags, dates)
 
 
-def _persist_to_db(site, kind):
+def _persist_to_db(site, kind, force=False, target_date=None):
     """下载成功后把本次数据入库 PostgreSQL（尽力而为，绝不外抛，不影响任务判定）。
     - __compare__ 无源数据，跳过。
     - auto_ingest=false 时跳过（无 PG/cpolar 的开发机）。
+    - force + target_date：应到类先按 business_date 清空本次下载日再整批重入（防反复入库归属漂移）。
     - 懒导入 store 以回避 import 顺序（store↔compare 与 runtime↔compare 共存）。
     - 结果写 state_store.ingest_state，供 /api/status 反映入库健康。
     所有写库/写状态都包 try/except：失败仅告警，绝不改变 dispatch_task 的 SUCCESS 判定。"""
@@ -711,7 +712,7 @@ def _persist_to_db(site, kind):
         if not store.ingest_enabled():  # 移入 try：config.yaml 缺失/损坏时也不外抛
             print(">> [入库] 已关闭 (auto_ingest=false)，跳过")
             return
-        count = store.ingest_task(site, kind)
+        count = store.ingest_task(site, kind, force=force, target_date=target_date)
         # 4 站 undelivered 连带入了 expected+actual：按实际入库的类补记 ingest_state，
         # 否则心跳派生 ready（expected ∧ actual → undelivered）会读到陈旧值。
         logged = (
@@ -755,7 +756,12 @@ def dispatch_task(ctx, task_spec):
         if ret is False:
             return (state_store.TASK_FAILED, "任务执行失败（重试耗尽）")
         _record_business_date(site, kind, task_spec.get("date"))
-        _persist_to_db(site, kind)
+        _persist_to_db(
+            site,
+            kind,
+            force=bool(task_spec.get("force", False)),
+            target_date=task_spec.get("date"),
+        )
         return (state_store.TASK_SUCCESS, None)
     except Exception as e:
         return (state_store.TASK_FAILED, str(e))