The repository already included benchmark baselines, regression thresholds, and a bench:check script — but the CI workflow never actually executed benchmark checks. This meant performance regressions could silently pass even though the documentation described CI benchmark enforcement. This PR aligned the actual CI behavior with the documented benchmark workflow.
Changes included adding a dedicated benchmark job to .github/workflows/ci.yml that runs bun run bench:check, configuring it in advisory mode with continue-on-error: true to avoid flaky benchmark variance blocking merges initially, adding GitHub Actions ::warning annotations in benchmarks/check-regression.ts that include scenario name, metric name, baseline and current values, and percent regression, and updating benchmarks/README.md to document the advisory CI behavior.