feat(harness): 迁移 gate 信息量阶梯到 unit 粒度 + schema_version

核心算法保真#5（信息阶梯）：gate_ladder.py 从逐题迁移到 unit 粒度，
只换键 question_id→unit_id，冷启动 2:1 错优先交错、gamma-EMA 公式、
Beta(1,1) 先验、反泄漏 _gate_ 过滤的公式/比例/顺序语义一字不改。

- LadderEntry 按 unit_id 键；AR pair 折叠为一个阶梯单元
- build_cold_entries 收单元列表，unit 错 = 任一成员错（双向 AND）折叠，
  2:1 交错 + probe 探针按 unit 抽，Beta 先验 p0 不变
- ladder_for 返回 unit_id 序、exclude 迁到 unit 口径（防半 pair 灌入
  触发下游 _ladder_units fail-fast）
- update_probs 先把逐题观测折叠成单元观测再按 unit_id 匹配更新，
  半观测单元跳过（防按 qid 匹配 pair 失效致 gamma-EMA 停摆）
- GatePools.save/load 加 schema_version=2；存量无版本/旧版本 json
  加载直接报错，拒绝静默混用 qid/unit 键
- BaselineCache 第四维键改名 unit_id（与 T7 validate 路径对齐）
- build_or_load_gate_pools 先折叠单元再排除 test（抽 helper 控复杂度 B）
- runner：_init_gate_pools 建 unit 索引；gate 验证 exclude/展开、
  _refresh_gate_ladder 折叠观测走 units_by_id

反泄漏 run_id 含 _gate_ 过滤不受影响（未改）。
测试：新增 test_gate_ladder_unit_migration.py（15 例覆盖 a-e），
既有 test_harness_gate_ladder.py 迁移到 unit API。全量 1363 passed。

app/harness/runner.py

@@ -854,6 +854,11 @@ class Runner:
         self._gate_questions_by_id: dict[str, GeneratedQuestion] = {
             q.question_id: q for q in questions
         }
+        # unit 索引：ladder_for 返回 unit_id、update_probs 折叠逐题观测均需按 unit_id
+        # 反查成员题（核心算法保真 #5：gate 阶梯 unit 化）。不进 checkpoint、每次启动重建。
+        self._gate_units_by_id: dict[str, QuestionUnit] = {
+            u.unit_id: u for u in build_units(questions)
+        }
         with HarnessLog(str(self._paths.db_path), pools.baseline_run_id) as log:
             rows = log.query(
                 "SELECT question_id, prediction, answer FROM predictions WHERE run_id=?",
@@ -1108,21 +1113,29 @@ class Runner:
         from app.harness.log import HarnessLog
         from app.harness.validate import validate_skill_local
 
-        exclude_qids = {c.question_id for c in pack.failure_cases + pack.success_cases}
-        ladder_qids = state.gate_pools.ladder_for(
+        # 案例包按 unit 排除：把每个 case 的 question_id 映射到其所属 unit_id，
+        # 命中单元整体排除，防止只排 AR pair 半个成员而给 gate 池灌半个 pair
+        # （下游 _ladder_units 会 fail-fast）。核心算法保真 #5。
+        exclude_units = {
+            self._gate_questions_by_id[c.question_id].unit_id
+            for c in pack.failure_cases + pack.success_cases
+            if c.question_id in self._gate_questions_by_id
+        }
+        ladder_unit_ids = state.gate_pools.ladder_for(
             task_type,
-            exclude_qids,
+            exclude_units,
             p_low=self._config.gate_p_low,
             p_high=self._config.gate_p_high,
             cold=not state.gate_epoch_observed,
         )
-        missing = [qid for qid in ladder_qids if qid not in self._gate_questions_by_id]
+        missing = [uid for uid in ladder_unit_ids if uid not in self._gate_units_by_id]
         if missing:
             raise ValueError(
-                f"gate 阶梯[{task_type}] 含 benchmark 中不存在的题: "
+                f"gate 阶梯[{task_type}] 含 benchmark 中不存在的单元: "
                 f"{missing[:5]}（gate_pools.json 与题库失配）"
             )
-        ladder_items = [self._gate_questions_by_id[qid] for qid in ladder_qids]
+        # 单元展开为逐题（unit 内成员顺序保持），下游 validate 再按阶梯序聚合回单元。
+        ladder_items = [q for uid in ladder_unit_ids for q in self._gate_units_by_id[uid].questions]
         base_skill_content = (self._paths.skills_dir / record.target_file).read_text(
             encoding="utf-8"
         )
@@ -1799,7 +1812,10 @@ class Runner:
         obs.update({r["question_id"]: r["prediction"] == r["answer"] for r in slow_rows})
         for extra_rows in extra_rows_lists:
             obs.update({r["question_id"]: r["prediction"] == r["answer"] for r in extra_rows})
-        state.gate_pools.update_probs(obs, gamma=self._config.gate_gamma_decay)
+        # 逐题观测折叠成单元观测后按 unit_id 匹配更新（防按 qid 匹配 pair 失效致 EMA 停摆）。
+        state.gate_pools.update_probs(
+            obs, self._gate_units_by_id, gamma=self._config.gate_gamma_decay
+        )
         state.gate_pools.save(self._config.workspace_dir / "gate_pools.json")
         state.gate_epoch_observed = True