Promote llama.cpp and LangGraph posture

[8Dionysus]

Summary

• record llama.cpp as the preferred bounded local-worker runtime and Ollama as the control path
• record LangGraph as the adopted bounded execution layer for W5/W6 style flows
• move the general Qwen helper defaults to the promoted 5403 path while keeping 5401 as the explicit control path

Validation

• python3 -m py_compile scripts/aoa-machine-fit scripts/aoa-qwen-run scripts/aoa-qwen-check
• bash -n scripts/aoa-qwen-bench
• python3 scripts/validate_stack.py
• curl -fsS http://127.0.0.1:5403/health
• curl -fsS http://127.0.0.1:5401/health
• scripts/aoa-qwen-check --case exact-reply --json