chore(bench): add cbm-vs-graphify benchmark suite

[yfjin-png]

Summary

codebase-memory (cbm) と graphify を比較するベンチマーク一式を bench/ 配下に追加します。

内容

• スクリプト: build_oracle.py, generate_qbank.py, compare_layer1.py, measure_layer3.py, analyze_layer2.py, qa_workflow.js, headtohead.sh
• 結果: oracle.json, layer2_full.json, qbank.jsonl, summary.json ほか (bench/cbm-vs-graphify/results/)
• レポート: REPORT_cbm_vs_graphify.md, BASELINE.md, README.md

19 files changed, 2015 insertions(+).

Notes

ドキュメント/ベンチマークの追加のみで、製品コードへの変更はありません。

[DeusData]

Huge thanks for opening this PR and for the work you put into it.

The maintainer shop is currently full, so this may sit for a bit before it gets a proper review. We will come back to this as soon as possible with real feedback; I wanted to make sure it did not sit unacknowledged in the meantime.

[DeusData]

Thank you for putting this together — head-to-head benchmarking against graphify is genuinely useful signal. We're going to close this one, for a couple of practical reasons rather than the idea: the DCO check is failing (sign-off email mismatch), and the suite bakes in machine-specific state — hardcoded absolute paths (/Users/charlesqin/…, /Users/yf.jin/.local/bin/…) and ~1900 lines of committed result JSON — so it isn't reproducible for anyone else as-is. If you'd like to bring it back as portable tooling (parameterized binary paths + repo root, results out of git, DCO signed-off), we'd be glad to take another look. Appreciate the effort and the competitive analysis!