feat(worker): 手动查重 worker task + visual_similarity/match_count 字段 #1661

- 新增 Celery task worker.process_duplication_check：
  OSS 下载上传视频 → 动态抽帧指纹 → 跨项目与用户所有成片比对
  (compute_duplicate_rate + find_duplicate_segments) → 更新 DuplicationRecord
- 状态流转 pending → processing → completed/failed，失败重试 3 次间隔 60s，
  重试耗尽标记 failed，临时文件 finally 清理
- 片段映射：worker 分片 DuplicateSegment(ms/汉明距离) → 领域 DuplicateSegment(秒/0-100 相似度)
- DuplicationRecord 实体新增 visual_similarity/match_count（可选 None，向后兼容），
  mark_completed()/reset_for_retry() 同步
- migration 065: duplication_records 表加两列
- repository/model/schemas/routes 读写映射与响应透传
- 上传接口与重试接口 enqueue task（此前记录一直 pending 的根因）
- 11 个新单测：成功/跳过/未找到/重试/失败标记/临时文件清理/片段映射/schema

packages/adapters/sqlalchemy_impl/duplication_repository.py

@@ -25,6 +25,8 @@ class SQLAlchemyDuplicationRecordRepository:
             status=record.status,
             duplicate_rate=record.duplicate_rate,
             duplicate_count=record.duplicate_count,
+            visual_similarity=record.visual_similarity,
+            match_count=record.match_count,
             video_fingerprint=json.dumps(record.video_fingerprint) if record.video_fingerprint else None,
             error_message=record.error_message,
             created_at=record.created_at,
@@ -58,6 +60,8 @@ class SQLAlchemyDuplicationRecordRepository:
         model.status = record.status
         model.duplicate_rate = record.duplicate_rate
         model.duplicate_count = record.duplicate_count
+        model.visual_similarity = record.visual_similarity
+        model.match_count = record.match_count
         model.video_fingerprint = json.dumps(record.video_fingerprint) if record.video_fingerprint else None
         model.error_message = record.error_message
         model.updated_at = record.updated_at
@@ -121,6 +125,8 @@ class SQLAlchemyDuplicationRecordRepository:
             status=model.status,
             duplicate_rate=model.duplicate_rate,
             duplicate_count=int(model.duplicate_count or 0),
+            visual_similarity=getattr(model, "visual_similarity", None),
+            match_count=getattr(model, "match_count", None),
             video_fingerprint=json.loads(fp_raw) if fp_raw else None,
             error_message=getattr(model, "error_message", ""),
             segments=segments,

packages/adapters/sqlalchemy_impl/models.py

@@ -417,6 +417,9 @@ class DuplicationRecordModel(Base):
     status = Column(String(20), nullable=False, default="pending", index=True)
     duplicate_rate = Column(Float, nullable=True)
     duplicate_count = Column(Integer, nullable=False, default=0)
+    # #1661 手动查重：视觉相似度（0~1）/ 匹配视频数
+    visual_similarity = Column(Float, nullable=True)
+    match_count = Column(Integer, nullable=True)
     video_fingerprint = Column(Text, nullable=True)
     error_message = Column(Text, nullable=False, default="")
     created_at = Column(DateTime, nullable=False, default=lambda: datetime.now(timezone.utc))