fix(inc-20260425): A1 三段 Agent timeout 拆分 + A2 DIAGNOSE 移除 Ollama

INC-20260425-8D17BB / 3B6C39 兩則告警 AI 信心降到 20% 根因雙修（統帥批准 A+B）：

A1 — 三段 Agent step timeout 拆分（北極星 §1.2 Observable by Default）:
- diagnostician_agent.py: PHASE2_STEP_TIMEOUT_SEC=20.0 共用值 → 拆三段
  · AGENT_DIAGNOSTICIAN_TIMEOUT_SEC=30.0（NIM 主吃口，最大 prompt + 多假設）
  · AGENT_SOLVER_TIMEOUT_SEC=20.0（後續 commit 接線）
  · AGENT_CRITIC_TIMEOUT_SEC=15.0（後續 commit 接線）
  · env override 支援，K8s ConfigMap 動態調整不需 rebuild
  · 保留 PHASE2_STEP_TIMEOUT_SEC alias（DEPRECATED，下 sprint 移除）
- observability/agent_step_metrics.py (58 行) — 新模組:
  · aiops_agent_step_duration_seconds Histogram
  · observe_agent_step() helper 統一三 Agent 呼叫點
  · outcome label ∈ {success, timeout, error}
  · agent label ∈ {diagnostician, solver, critic}

A2 — ai_router DIAGNOSE chain 移除 Ollama:
- ai_router.py v4.4 by Claude Sonnet 4.6
  · 新增 _diagnose_fallback_chain: NEMO → GEMINI → CLAUDE
  · Ollama 永久排除於此 chain（CPU-only 實測 238s，二次 timeout 必爆）
  · 新增 aiops_diagnose_fallback_total Prometheus metric
- 根因: NIM timeout 後 fallback 到 Ollama deepseek-r1:14b CPU 238s
  → 二次 timeout → degraded confidence=0.2

Wave8-X2 整合測試補正:
- test_ollama_failover_manager.py: TestSelectProvider 補 mock _check_gemini_quota
  原 test 期望 OFFLINE→Gemini，但 quota fail-closed 後沒 mock 會被切到 188
  繞過 quota check 後驗純路由邏輯 → 37/37 PASS

Tests: 37 passed (test_ollama_failover_manager 全部)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 (Wave 8 INC-20260425) <noreply@anthropic.com>

apps/api/tests/test_ollama_failover_manager.py

@@ -331,11 +331,21 @@ class TestSelectProvider:
         manager = OllamaFailoverManager(health_monitor=mock_monitor)
         manager._settings = mock_settings
 
-        with patch.object(manager, "_write_failover_audit", return_value=None):
+        # 2026-04-27 Wave8-X2: 必須 mock Redis pool（_check_gemini_quota 走 fail-closed 路徑會切到 188 而非 Gemini）
+        # 測試本身只要驗 OFFLINE → Gemini 路由邏輯，故繞過 quota check
+        with patch.object(manager, "_write_failover_audit", return_value=None), \
+             patch.object(manager, "_check_gemini_quota", AsyncMock(return_value=True)), \
+             patch(
+                 "src.services.failover_alerter.get_failover_alerter",
+                 return_value=MagicMock(
+                     alert_failover=AsyncMock(),
+                     alert_gemini_quota_exceeded=AsyncMock(),
+                 ),
+             ):
             result = await manager.select_provider()
 
         assert isinstance(result, OllamaRoutingResult)
-        # 新矩陣：111 OFFLINE → primary=Gemini（188 降為 fallback）
+        # 新矩陣：111 OFFLINE + Gemini quota OK → primary=Gemini（188 降為 fallback）
         assert result.primary.provider_name == "gemini"
 
     @pytest.mark.asyncio
@@ -655,17 +665,21 @@ class TestGeminiQuota:
         assert ok is False
 
     @pytest.mark.asyncio
-    async def test_gemini_quota_redis_unavailable_fail_open(self):
-        """Redis 掛掉 → 返回 True（fail-open，仍允許走 Gemini）"""
+    async def test_gemini_quota_redis_unavailable_fail_closed(self):
+        """Redis 掛掉 → 返回 False（2026-04-27 Wave8-X2 fail-closed，違反費用鐵律的修復）"""
         manager = _make_manager()
 
         with patch(
             "src.core.redis_client.get_redis",
             side_effect=RuntimeError("Redis unavailable"),
+        ), patch(
+            "src.services.failover_alerter.get_failover_alerter",
+            return_value=MagicMock(alert_gemini_quota_exceeded=AsyncMock()),
         ):
             ok = await manager._check_gemini_quota()
 
-        assert ok is True
+        # fail-closed：Redis 異常時拒絕 Gemini，避免費用失控（违反 feedback_cost_change_approval.md）
+        assert ok is False
 
     @pytest.mark.asyncio
     async def test_select_provider_quota_exceeded_uses_188(self):