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

config/default.yaml

@@ -22,37 +22,6 @@ embed:
   embed_dim: 768
   device: "cpu"
 
-# ── 可训练检索器 ──
-retriever:
-  embed_dim: 768
-  num_heads: 4
-  L_layers: 2
-  L_cycles: 4
-  max_rounds: 5
-  ffn_expansion: 2.0
-  checkpoint: null
-  k_l1: 1
-  k_l2: 1
-  k_l3: 1
-  max_paths: 5
-
-# ── 检索器训练 ──
-train:
-  lr: 1.0e-4
-  weight_decay: 1.0e-5
-  batch_size: 1
-  max_epochs_phase1: 30
-  max_epochs_phase2: 20
-  nav_loss_weight: 1.0
-  act_loss_weight: 0.1
-  margin_loss_weight: 0.5
-  act_lambda_step: 0.1
-  act_gamma: 0.9
-  eval_interval: 5
-  save_dir: "checkpoints"
-  dataset: "videomme"
-  dataset_path: "data/videomme/splits/train.jsonl"
-
 # ── Harness 自进化循环 ──
 harness:
   workspace_dir: "workspaces/default"