存量积累：生图素人感约束+评图分+幂等防重跑+审核回路

- 015-017迁移：image_candidate 文案复审/AI视觉分/重生标记
- constants C7素人感约束(反电商摆拍对齐真实笔记)+C3叠字口子
- celery visibility_timeout=2h 防长任务被误判重投重复烧钱(task75教训)
- image_scorer 评图分(只筛选+展示,真实信号才进飞轮权重)
- storyboard/image_gen/pipeline_io 生图存量
- task_actions/tasks/task_service/flywheel 审核回路+飞轮存量

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

backend/app/api/v1/tasks.py

@@ -60,6 +60,10 @@ def _fmt_task(t: GenerationTask) -> dict:
         "id": t.id, "product_id": t.product_id, "theme": t.theme,
         "status": t.status, "text_count": t.text_count, "image_count": t.image_count,
         "track": t.track, "need_product_image": t.need_product_image,
+        # 审核回路：打回原因+审核状态,任务页/历史页展示"为何被打回"(R4-b死链修复,R8历史复用)
+        "review_status": t.review_status,
+        "reject_reason": t.reject_reason,
+        "reviewed_at": t.reviewed_at.isoformat() if t.reviewed_at else None,
         "created_at": t.created_at.isoformat(),
     }
 
@@ -129,6 +133,7 @@ def _fmt_text(tc: TextCandidate) -> dict:
         # AI 评委总评（verdict/summary 存在 content 列，新数据有，旧数据为空字符串降级）
         "verdict": parsed.get("verdict", ""),         # "优秀"|"合格"|"不合格"
         "summary": parsed.get("summary", ""),         # 一句话总评含改进点
+        "edited": tc.edited,                          # D1 改稿标记(飞轮最强信号)
     }
 
 
@@ -136,6 +141,9 @@ def _fmt_image(ic: ImageCandidate) -> dict:
     return {
         "candidate_id": ic.id, "role": ic.role, "url": ic.url,
         "strategy": ic.strategy, "seq": ic.seq, "is_selected": ic.is_selected,
+        "is_regen": getattr(ic, "is_regen", False),   # R2 重生标记(前端同strategy+role去重展示最新)
+        # E12 AI评图分：纯展示+排序,绝不进飞轮权重(eval_score 另留 NULL)
+        "ai_visual_score": ic.ai_visual_score, "ai_visual_note": ic.ai_visual_note,
     }