feat: 添加 Topic-SubTopic 层级结构与 Accordion 侧边栏

- topics/index.json 重构为嵌套模型（topic → subtopics）
- 侧边栏改为 Accordion 交互：默认折叠、同时只展开一个
- 展开/折叠状态通过 localStorage 持久化
- 更新架构文档同步数据模型变更

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

topics/index.json

@@ -3,69 +3,76 @@
   "updated": "2026-09-02",
   "topics": [
     {
-      "slug": "gc-jvm",
-      "name": "GC / JVM / 三色标记",
-      "description": "JVM 堆内存模型、GC 算法、三色标记法、垃圾收集器",
-      "path": "topics/gc-jvm",
-      "stats": {
-        "total": 20,
-        "by_type": {
-          "fill_blank": 10,
-          "single_choice": 10
+      "slug": "qunar-ai-fullstack",
+      "name": "去哪儿 AI 面 — AI 全栈方向",
+      "description": "覆盖 JVM/GC、AI Agent 架构、工具链、幻觉治理等核心考察领域",
+      "subtopics": [
+        {
+          "slug": "gc-jvm",
+          "name": "GC / JVM / 三色标记",
+          "description": "JVM 堆内存模型、GC 算法、三色标记法、垃圾收集器",
+          "path": "topics/gc-jvm",
+          "stats": {
+            "total": 20,
+            "by_type": {
+              "fill_blank": 10,
+              "single_choice": 10
+            }
+          }
+        },
+        {
+          "slug": "ai-structured-output",
+          "name": "AI 结构化输出 / Agent 架构",
+          "description": "结构化输出方法、Agent 模式、受限解码",
+          "path": "topics/ai-structured-output",
+          "stats": {
+            "total": 20,
+            "by_type": {
+              "fill_blank": 10,
+              "single_choice": 10
+            }
+          }
+        },
+        {
+          "slug": "ai-memory",
+          "name": "Agent 记忆管理 / 冲突处理",
+          "description": "记忆分层架构、淘汰策略、冲突检测与消解",
+          "path": "topics/ai-memory",
+          "stats": {
+            "total": 20,
+            "by_type": {
+              "fill_blank": 10,
+              "single_choice": 10
+            }
+          }
+        },
+        {
+          "slug": "ai-hallucination",
+          "name": "AI 幻觉 / 目标评价 / 规范",
+          "description": "幻觉分类、Critic 架构、Context Engineering、评估体系",
+          "path": "topics/ai-hallucination",
+          "stats": {
+            "total": 20,
+            "by_type": {
+              "fill_blank": 10,
+              "single_choice": 10
+            }
+          }
+        },
+        {
+          "slug": "ai-tooling",
+          "name": "AI 工具 / Harness / 综合",
+          "description": "AI 工具选型、Harness 架构、可靠性保障、提效分析",
+          "path": "topics/ai-tooling",
+          "stats": {
+            "total": 20,
+            "by_type": {
+              "fill_blank": 10,
+              "single_choice": 10
+            }
+          }
         }
-      }
-    },
-    {
-      "slug": "ai-structured-output",
-      "name": "AI 结构化输出 / Agent 架构",
-      "description": "结构化输出方法、Agent 模式、受限解码",
-      "path": "topics/ai-structured-output",
-      "stats": {
-        "total": 20,
-        "by_type": {
-          "fill_blank": 10,
-          "single_choice": 10
-        }
-      }
-    },
-    {
-      "slug": "ai-memory",
-      "name": "Agent 记忆管理 / 冲突处理",
-      "description": "记忆分层架构、淘汰策略、冲突检测与消解",
-      "path": "topics/ai-memory",
-      "stats": {
-        "total": 20,
-        "by_type": {
-          "fill_blank": 10,
-          "single_choice": 10
-        }
-      }
-    },
-    {
-      "slug": "ai-hallucination",
-      "name": "AI 幻觉 / 目标评价 / 规范",
-      "description": "幻觉分类、Critic 架构、Context Engineering、评估体系",
-      "path": "topics/ai-hallucination",
-      "stats": {
-        "total": 20,
-        "by_type": {
-          "fill_blank": 10,
-          "single_choice": 10
-        }
-      }
-    },
-    {
-      "slug": "ai-tooling",
-      "name": "AI 工具 / Harness / 综合",
-      "description": "AI 工具选型、Harness 架构、可靠性保障、提效分析",
-      "path": "topics/ai-tooling",
-      "stats": {
-        "total": 20,
-        "by_type": {
-          "fill_blank": 10,
-          "single_choice": 10
-        }
-      }
+      ]
     }
   ]
-}
+}