refactor(lyrics-cache): 重构歌词缓存读取写入逻辑并同步新增歌词缓存

- 抽取歌词缓存读取和写入为独立函数 _read_existing_lyrics_cache 与 _write_existing_lyrics_cache
- load_existing_l2_candidates 中改用新缓存读写函数，增加缓存加载日志和进度输出
- 处理下载异常时保证已下载缓存的歌词及时持久化，避免丢失数据
- 新增 sync_inserted_lyrics_cache 函数实现新增歌词缓存同步写入功能
- 后台写入线程写入数据库后调用歌词缓存同步，记录新增条数日志
- 修改批量写入队列数据结构，传递缓存同步相关数据
- 优化去重统计日志显示格式，新增 format_dedup_stats 函数统一格式化
- L2审核界面新增 skip 筛选选项，调整相关前端和后端代码支持该状态
- 移除 L2 审核界面中保存按钮，避免误操作手动保存
- 补充单元测试覆盖缓存同步、缓存读取异常处理和去重统计日志格式输出

@@ -466,6 +466,7 @@
         <option value="new">new</option>
         <option value="new">new</option>
         <option value="merge">merge</option>
         <option value="merge">merge</option>
         <option value="review">review</option>
         <option value="review">review</option>
+        <option value="skip">skip</option>
       </select>
       </select>
       <input id="searchInput" type="search" placeholder="搜索歌名 / ID">
       <input id="searchInput" type="search" placeholder="搜索歌名 / ID">
       <label for="pageSizeSelect">每页</label>
       <label for="pageSizeSelect">每页</label>
@@ -858,7 +859,6 @@
                     ${value === 'duplicate' ? '确认重复' : value === 'not_duplicate' ? '确认不重复' : '待确认'}
                     ${value === 'duplicate' ? '确认重复' : value === 'not_duplicate' ? '确认不重复' : '待确认'}
                   </button>`).join('')}
                   </button>`).join('')}
                 <input id="reviewNote" value="${esc(selectedReview.note || '')}" placeholder="人工备注">
                 <input id="reviewNote" value="${esc(selectedReview.note || '')}" placeholder="人工备注">
-                <button id="saveReviewBtn" class="secondary">保存</button>
                 <button id="deleteReviewBtn" class="secondary" style="color:#c0392b;border-color:#c0392b">删除</button>
                 <button id="deleteReviewBtn" class="secondary" style="color:#c0392b;border-color:#c0392b">删除</button>
               </div>
               </div>
             `}
             `}
@@ -906,14 +906,6 @@
           renderDetail();
           renderDetail();
         });
         });
       }
       }
-      const saveBtn = document.getElementById('saveReviewBtn');
-      if (saveBtn) {
-        saveBtn.addEventListener('click', () => {
-          if (!candidate) return;
-          setReview(candidate, { note: document.getElementById('reviewNote')?.value || '' });
-          renderDetail();
-        });
-      }
       const deleteBtn = document.getElementById('deleteReviewBtn');
       const deleteBtn = document.getElementById('deleteReviewBtn');
       if (deleteBtn) {
       if (deleteBtn) {
         deleteBtn.addEventListener('click', async () => {
         deleteBtn.addEventListener('click', async () => {

@@ -174,8 +174,8 @@ def _staging_groups_response(mode: str, term: str, page: int, page_size: int, de
     # hits 模式下只展示有命中的记录（merge/review）
     # hits 模式下只展示有命中的记录（merge/review）
     if mode == "hits":
     if mode == "hits":
         where_clauses.append("dedup_action IN ('merge', 'review')")
         where_clauses.append("dedup_action IN ('merge', 'review')")
-    # 决策筛选：new / merge / review
-    if decision and decision in ('new', 'merge', 'review'):
+    # 决策筛选：new / merge / review / skip
+    if decision and decision in ('new', 'merge', 'review', 'skip'):
         where_clauses.append("dedup_action = %s")
         where_clauses.append("dedup_action = %s")
         params.append(decision)
         params.append(decision)
     where = "WHERE " + " AND ".join(where_clauses) if where_clauses else ""
     where = "WHERE " + " AND ".join(where_clauses) if where_clauses else ""
@@ -640,7 +640,7 @@ def _groups_response(path: Path, top_k: str, mode: str, term: str, page: int, pa
     if mode == "hits":
     if mode == "hits":
         groups = [g for g in groups if g.get("has_hit")]
         groups = [g for g in groups if g.get("has_hit")]
     # 决策筛选
     # 决策筛选
-    if decision and decision in ('new', 'merge', 'review', 'duplicate'):
+    if decision and decision in ('new', 'merge', 'review', 'duplicate', 'skip'):
         groups = [g for g in groups if g.get("decision") == decision or (decision == 'merge' and g.get("decision") == 'merge')]
         groups = [g for g in groups if g.get("decision") == decision or (decision == 'merge' and g.get("decision") == 'merge')]
     filtered = [group for group in groups if _matches_term(group, term)]
     filtered = [group for group in groups if _matches_term(group, term)]
     start = max(0, (page - 1) * page_size)
     start = max(0, (page - 1) * page_size)

@@ -43,6 +43,8 @@ from import_hk_songs import (
     process_lyrics,
     process_lyrics,
     DedupReport,
     DedupReport,
     L2CandidateIndex,
     L2CandidateIndex,
+    format_dedup_stats,
+    sync_inserted_lyrics_cache,
     load_approved_import_ids,
     load_approved_import_ids,
     load_existing_l2_candidates,
     load_existing_l2_candidates,
 )
 )
@@ -140,6 +142,26 @@ class TestPendingL1Index:
         assert len(l1_index) == 1
         assert len(l1_index) == 1
 
 
 
 
+class TestDedupStatsLog:
+    """测试去重统计日志文案"""
+
+    def test_l1_counter_is_labeled_as_match_not_skip(self):
+        stats = {
+            'l1_dup': 2,
+            'l2_dup': 2,
+            'l2_review': 23,
+            'l2_new': 1725,
+            'author_merged': 1,
+            'skipped': 0,
+        }
+
+        message = format_dedup_stats(stats)
+
+        assert 'L1命中=2' in message
+        assert 'L1跳过' not in message
+        assert '最终跳过=0' in message
+
+
 class TestDbWriterShutdown:
 class TestDbWriterShutdown:
     """测试后台 DB 写入线程的优雅停止"""
     """测试后台 DB 写入线程的优雅停止"""
 
 
@@ -213,6 +235,72 @@ class TestExistingLyricsCache:
         assert records[0].lyrics == '缓存里的歌词\n第二行'
         assert records[0].lyrics == '缓存里的歌词\n第二行'
         assert stats == {'cached': 1, 'downloaded': 1, 'failed': 0}
         assert stats == {'cached': 1, 'downloaded': 1, 'failed': 0}
 
 
+    def test_load_existing_l2_candidates_persists_successes_before_later_download_exception(self, tmp_path):
+        cache_path = tmp_path / 'lyrics_cache.jsonl'
+        rows = [
+            {
+                'id': 100,
+                'lyrics_url': 'https://oss.example.test/100.txt',
+                'name': '第一首',
+                'singer': '歌手',
+                'lyricist': '词',
+                'composer': '曲',
+            },
+            {
+                'id': 101,
+                'lyrics_url': 'https://oss.example.test/101.txt',
+                'name': '第二首',
+                'singer': '歌手',
+                'lyricist': '词',
+                'composer': '曲',
+            },
+        ]
+
+        def downloader(url, timeout=10):
+            if url.endswith('/101.txt'):
+                raise RuntimeError('network interrupted')
+            return '已下载歌词'.encode('utf-8'), 'text/plain'
+
+        with pytest.raises(RuntimeError, match='network interrupted'):
+            load_existing_l2_candidates(rows, cache_path, downloader=downloader)
+
+        cache_lines = cache_path.read_text(encoding='utf-8').splitlines()
+        assert len(cache_lines) == 1
+        cached = json.loads(cache_lines[0])
+        assert cached['id'] == '100'
+        assert cached['lyrics'] == '已下载歌词'
+
+    def test_sync_inserted_lyrics_cache_adds_committed_inserted_rows(self, tmp_path):
+        cache_path = tmp_path / 'lyrics_cache.jsonl'
+        cache_path.write_text(
+            json.dumps({
+                'id': '100',
+                'url': 'https://oss.example.test/100.txt',
+                'lyrics': '已有歌词',
+            }, ensure_ascii=False) + '\n',
+            encoding='utf-8',
+        )
+        tuple_row = [None] * 45
+        tuple_row[0] = 200
+        tuple_row[1] = '新歌'
+        tuple_row[2] = '新词'
+        tuple_row[3] = '新曲'
+        tuple_row[8] = 'https://oss.example.test/200.txt'
+        tuple_row[33] = '新歌手'
+        source_row = {'lyrics_txt_content': '新歌词正文'}
+
+        added = sync_inserted_lyrics_cache(cache_path, [(tuple(tuple_row), source_row)])
+
+        cache_items = [
+            json.loads(line)
+            for line in cache_path.read_text(encoding='utf-8').splitlines()
+        ]
+        assert added == 1
+        assert len(cache_items) == 2
+        assert cache_items[1]['id'] == '200'
+        assert cache_items[1]['url'] == 'https://oss.example.test/200.txt'
+        assert cache_items[1]['lyrics'] == '新歌词正文'
+
 
 
 # ====================================================================
 # ====================================================================
 # L1 元数据去重 check_l1
 # L1 元数据去重 check_l1