revert 3bafd30aad

revert feat(HEL-529): 按定稿100%重做三页数据中枢 + 数据修正1-5

视觉：admin/app.js 按已确认打样 hub-kimi.html 逐行重写三页 DOM 与动效
（sparkline/缓存年龄秒增/延迟变化闪烁/EVENT TAPE 预装滚动/分组接口表/
更新频率列/时钟冒号 blink/雷达 blip），CSS 补真实调用脉冲 node-ping。
数据修正：
1) pipeline._stage 批内按暂存表业务键确定性去重（保留最后一条），
   修复人气榜/龙虎榜自 09-07 起每日 UNIQUE constraint 落库失败；
2) overview.anomalies 收敛为「最新批次未成功且当日未发布」的当前异常，
   历史已恢复批次留在审计明细；
3) source_catalog 接口补真实批次分组 + 观测 join（接口名或数据集名
   双向匹配，标注 observed/observed_basis），消除「全部未配置/0/30」误报；
4) lineage 逐数据集按其服务接口过滤健康行（接口级状态），
   provisional 已配置无观测显示「已配置 · 待观测」，仅 iFinD 为未配置；
5) lineage 补 update_freq 真实频率字段。
测试：新增 tests/test_hel529_rework.py（8 项），全套 191 项通过
（1 项环境依赖失败在基线 d9358ab 上同样复现，与本改动无关）。

Co-authored-by: multica-agent <github@multica.ai>

xiaobai-datahub/datahub/pipeline.py

@@ -99,58 +99,6 @@ STAGING_INSERT = {
     **EXTENDED_STAGING_INSERT,
 }
 
-# Staging-table business keys (PRIMARY KEY minus the constant batch_id),
-# mirroring the PRIMARY KEY clauses declared in datahub/db.py and
-# datahub/datasets_ext.py. Used only to collapse within-batch duplicates so
-# a single upstream response cannot fail the whole batch on a UNIQUE
-# constraint (HEL-529: 2026-09-14 dc_hot returned 4 duplicate ts_codes and
-# hm_detail 43 duplicate (ts_code, hm_name) keys in one response, which has
-# blocked popularity/dragon_tiger publishing every day since 09-07).
-STAGING_KEY_FIELDS = {
-    "stocks": ("ts_code", "trade_date"),
-    "daily": ("ts_code", "trade_date"),
-    "valuation": ("ts_code", "trade_date"),
-    "moneyflow": ("ts_code", "trade_date"),
-    "auction": ("ts_code", "trade_date"),
-    "index_daily": ("ts_code", "trade_date"),
-    "limit_events": ("ts_code", "trade_date", "limit_type"),
-    "popularity": ("ts_code", "trade_date", "source"),
-    "dragon_tiger": ("ts_code", "trade_date", "hm_name"),
-    "sector_daily": ("ts_code", "trade_date", "family"),
-}
-
-
-def _dedupe_staging_rows(dataset: str, rows: list[dict[str, Any]]) -> list[dict[str, Any]]:
-    """Collapse within-batch duplicates on the staging table's business key.
-
-    Deterministic: keeps the LAST occurrence of each key (the same row the
-    eod copy's INSERT OR REPLACE would keep), preserves first-seen order, and
-    never touches rows across batches. Datasets without a declared business
-    key are returned unchanged.
-    """
-    fields = STAGING_KEY_FIELDS.get(dataset)
-    if not fields:
-        return rows
-    seen: dict[tuple, int] = {}
-    out: list[dict[str, Any]] = []
-    dropped = 0
-    for row in rows:
-        key = tuple(row.get(field) for field in fields)
-        if key in seen:
-            out[seen[key]] = row
-            dropped += 1
-        else:
-            seen[key] = len(out)
-            out.append(row)
-    if dropped:
-        LOGGER.info(
-            "staging dedupe: %s collapsed %d duplicate rows within batch (kept last)",
-            dataset,
-            dropped,
-            extra={"hub": {"dataset": dataset, "deduped": dropped}},
-        )
-    return out
-
 EOD_COPY = {
     "stocks": (
         "INSERT OR REPLACE INTO eod_stocks "
@@ -1772,7 +1720,6 @@ class Pipeline:
 
     def _stage(self, dataset: str, batch_id: str, rows: list[dict[str, Any]]) -> None:
         sql, mapper = STAGING_INSERT[dataset]
-        rows = _dedupe_staging_rows(dataset, rows)
         with self.db.write() as connection:
             connection.execute(
                 f"DELETE FROM {DATASET_TABLES[dataset][1]} WHERE batch_id = ?",