fix(data): corpus 全 fetcher 补 UA+jsonp 解析防御 [nas]

NAS 冒烟二跑实锤: search-api 对无 UA 请求返 200+空 body(金丝雀当时带 UA 故
未暴露)→s.index('(') ValueError 裸崩。修: 统一 UA 常量六处 fetcher 全带;
jsonp 非 cb(...) 形态 → TransportError(带响应片段,计入域名连续错走冷却
而非裸崩);容器内 A/B 实测 no-UA='' vs withUA=正常610条

scripts/data_platform/corpus_download.py

@@ -78,6 +78,7 @@ PDF_CAPACITY_CHECK_GB = 50      # 每落 50GB 自查一次 df
 
 RATE = {"cninfo": 1.0, "eastmoney": 1.0, "article": 2.0}   # 每域最小间隔(秒), 慢爬纪律
 JITTER = (0.0, 0.3)
+UA = {"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36"}
 
 ID_KEYS = {"ann_meta": ["announcement_id"],
            "news_meta": ["art_code", "stock_code"],
@@ -196,7 +197,7 @@ def load_stock_pool(refresh=False):
             j = json.loads(cache.read_text(encoding="utf-8"))
             return sorted(j.items())
     client = DomainClient("cninfo")
-    r = client.request("GET", POOL_URL)
+    r = client.request("GET", POOL_URL, headers=UA)
     pairs = [(it["code"], it["orgId"])
              for it in r.json().get("stockList") or []]
     pairs = filter_stock_pool(pairs)
@@ -424,7 +425,7 @@ def fetch_cninfo_announcements(client, code, org_id, start, end, category=""):
     out, page = [], 1
     while True:
         data["pageNum"] = str(page)
-        r = client.request("POST", QUERY_URL, data=data)
+        r = client.request("POST", QUERY_URL, data=data, headers=UA)
         try:
             j = r.json()
         except ValueError as e:
@@ -446,24 +447,26 @@ def fetch_news_recent(client, code):
                                             "postTag": "</em>"}}}
     params = {"cb": "cb", "param": json.dumps(inner, ensure_ascii=False),
               "_": str(int(time.time() * 1000))}
-    r = client.request("GET", SEARCH_API_URL, params=params)
+    r = client.request("GET", SEARCH_API_URL, params=params, headers=UA)
     s = r.text.strip()
+    if "(" not in s or not s.endswith(")"):
+        raise TransportError(f"search-api 非 jsonp 响应: {s[:120]!r}")
     j = json.loads(s[s.index("(") + 1:s.rindex(")")])
     return (j.get("result") or {}).get("cmsArticleWebOld") or []
 
 
 def fetch_news_backfill_page(client, code, mkt_prefix, page):
     url = NP_LIST_URL.format(mkt=mkt_prefix, code=code, page=page)
-    r = client.request("GET", url)
+    r = client.request("GET", url, headers=UA)
     return (r.json().get("data") or {}).get("list") or []
 
 
 def fetch_article_html(client, url):
-    return client.request("GET", url).text
+    return client.request("GET", url, headers=UA).text
 
 
 def download_pdf(client, url, dest):
-    r = client.request("GET", url)
+    r = client.request("GET", url, headers=UA)
     dest.parent.mkdir(parents=True, exist_ok=True)
     tmp = dest.with_suffix(".tmp")
     tmp.write_bytes(r.content)