feat(server): 参考 BidEngine 补五项：原因码、截断重跑、项目信息逐项表态、占位符、提示词规则对应校验

- 待确认记原因码（extract.FAILURES，存在 ai_values.failures），解析结果按原因计数
  （verify_failures）；前端待确认条目显示具体原因
- 相邻分块重叠约 800 字；模型输出被截断或不是合法 JSON 时对半重跑（最多 3 层），
  拆到底仍不完整记 output_incomplete。修复：以前截断时整块条目被当成「没有结果」悄悄丢掉
- 项目基本信息提示词 v2：12 个字段逐项表态 found / ambiguous / not_found，跨分块合并为
  found / absent / needs_review（info_fields）
- 「____元」「××」「（ ）」「待填」等占位符不算取值
- 提示词 substantive v4，qualification / scoring / schedule / tech_param v3：每条禁止事项
  对应一个原因码，对照表在 prompts/README.md
- 顾问（Fable）深读参考项目的结论补进 ref-bid-engine-analysis.md §5；架构 §13 SK12
- 待确认：国动 4 → 2，S02 13 → 3，S06 57 → 55

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

evals/run_extract.py

@@ -23,6 +23,7 @@ from typing import Any
 from uuid import UUID
 
 from bid_doctor.ai import prompts
+from bid_doctor.ai import service as ai_service
 from bid_doctor.ai.service import provider
 from bid_doctor.checklist import extract
 from bid_doctor.core.db import TenantContext
@@ -72,9 +73,12 @@ async def cached_run_json(ctx: Any, *, job_type: str, job_id: Any, task: str, pr
                     raise
                 print("retry", task, e, file=sys.stderr)
                 await asyncio.sleep(3)
-        data = {"data": r.data, "model": r.model, "tokens_in": r.tokens_in, "tokens_out": r.tokens_out}
+        data = {"data": r.data, "model": r.model, "tokens_in": r.tokens_in, "tokens_out": r.tokens_out,
+                "incomplete": r.incomplete}
         f.write_text(json.dumps(data, ensure_ascii=False))
     calls.append({"task": task, "version": version, "user": user, **data})
+    if data.get("incomplete"):
+        raise ai_service.OutputIncomplete(task)  # 和线上一样：输出被截断时由 extract.run 对半重跑
     return data["data"]
 
 
@@ -97,7 +101,7 @@ async def main() -> None:
         declared={m for n in notes for m in n.markers},
     )
     ctx = TenantContext(tenant_id=UUID(int=1), user_id=UUID(int=1))
-    items, info = await extract.run(ctx, UUID(int=2), doc)
+    items, info, run_stats = await extract.run(ctx, UUID(int=2), doc)
 
     by_sid = {f"p{s.page_no}s{s.seq}": s for s in body}
     # ---------- 结果 ----------
@@ -190,6 +194,9 @@ async def main() -> None:
         f"条目 {len(items)}：" + "，".join(f"{z}/{r or '-'} {n}" for (z, r), n in zone.most_common()),
         "按类别：" + "，".join(f"{c}/{z}:{n}" for (c, z), n in sorted(cats.items())),
         "项目信息：" + "，".join(f"{k}={getattr(v, 'value', v)}" for k, v in info.fields.items())[:600],
+        "项目信息结论：" + "，".join(f"{k}:{v['status']}{v.get('codes') or ''}" for k, v in info.status.items()),
+        "对半重跑 {} 次，拆到底仍不完整 {} 块".format(run_stats.get("chunks_split", 0), run_stats.get("output_incomplete", 0)),
+        "没通过校验的原因：" + "，".join(f"{k.removeprefix('fail:')} {n}" for k, n in sorted(run_stats.items()) if k.startswith("fail:")),
         "",
         "## 模型输出审计",
     ]