fix: 存库前先标记 keyword_matched，修复历史数据全为 false 的问题

pipeline 原来先存库再过滤，导致 keyword_matched 永远是 False。
改为先标记再存库，最新公告查询才能正确区分两个来源。

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app/services/pipeline.py

@@ -20,24 +20,28 @@ class PostCrawlPipeline:
         if config.dedup_enabled:
             announcements = dedup_by_hash(announcements)
 
-        # 2. Store to database
+        # 2. 先标记 keyword_matched，再存库
+        if config.filter_enabled and config.keywords:
+            for a in announcements:
+                a["keyword_matched"] = self._match_keywords(a, config.keywords)
+
+        # 3. Store to database
         stored = await self._save_to_db(announcements)
         result.stored = stored
 
         to_notify = announcements
 
-        # 3. Filter
+        # 4. Filter
         if config.filter_enabled and config.keywords:
             before = len(to_notify)
-            to_notify = [a for a in to_notify
+            to_notify = [a for a in to_notify if a.get("keyword_matched")]
-                         if self._match_keywords(a, config.keywords)]
             result.filtered = before - len(to_notify)
 
-        # 4. Skip already-notified
+        # 5. Skip already-notified
         if to_notify:
             to_notify = await self._exclude_sent(to_notify)
 
-        # 5. Notify
+        # 6. Notify
         if config.notify_mode == "all":
             result.notified = await self._send_notifications(to_notify)
         elif config.notify_mode == "filtered":
@@ -46,7 +50,7 @@ class PostCrawlPipeline:
             elif not config.filter_enabled:
                 result.notified = await self._send_notifications(to_notify)
 
-        # 6. Mark sent
+        # 7. Mark sent
         if config.mark_sent and result.notified > 0:
             await self._mark_sent(to_notify)