refactor: remove deprecated retriever module

RecursiveRetriever was a failed approach, not carried into TRM5.
- delete app/retriever/ (empty placeholder)
- drop retriever + train blocks from config/default.yaml
- renumber fidelity checklist 13->12 items (drop #4, shift up)
- sync core-goal text, dir tree, module-interaction diagrams
  across CLAUDE.md, ARCHITECTURE.md, overview.md, README.md
- reference/ kept intact as historical code

research-wiki/overview.md

@@ -1,6 +1,6 @@
 # 系统总览 (Overview)
 
-> Video-Tree-TRM5：在层次化视频树上构建可自我进化的搜索 Agent + 可训练递归检索器，通过 Harness Engineering 持续改进实现长视频理解。目标会议 EMNLP 2026。
+> Video-Tree-TRM5：在层次化视频树上构建可自我进化的搜索 Agent，通过 Harness Engineering 持续改进实现长视频理解。目标会议 EMNLP 2026。
 
 ## 1. 核心思想：自进化循环对标 PyTorch 训练
 
@@ -25,7 +25,6 @@ flowchart TD
     main[main.py CLI 入口] --> runner[app/harness/runner.py 训练循环]
     main --> build[app/tree/video_builder.py 建树]
     main --> qgen[app/question_gen/generator.py 新题构建]
-    main --> train_ret[app/retriever/train.py 检索器训练]
 
     runner --> inf[app/harness/inference.py 推理]
     runner --> diag[core/evolution/diagnose.py 诊断]
@@ -46,7 +45,6 @@ flowchart TD
 | `app/harness/` | 训练 harness：runner 循环编排、推理 step、mini-batch、信息阶梯、workspace 版本管理 |
 | `app/question_gen/` | 新题构建：题目生成、基线校准、去重 |
 | `app/search/` | 搜索 Agent 装配：PromptManager + SkillRegistry |
-| `app/retriever/` | 可训练检索器：RecursiveRetriever（CrossAttention+ACT）、两阶段训练 |
 | `core/agent/` | AgentLoop 引擎：Thinking+JSON 推理循环，pluggy hook 驱动 |
 | `core/evolution/` | 诊断+进化引擎：两阶段诊断、patch/rewrite 进化、CE-Gate e-process |
 | `adapters/` | 外部实现层：GovernedLLMClient（遥测+熔断+缓存）、VLM、Embedding、ASR、OCR |