perf: lazy load cli command runners

[freshtechbro]

Summary

• Lazy-load CLI command runner modules and keep help/version on cheap fast paths before command registration.
• Add a current-source CLI/tools latency benchmark harness and setup plan.
• Add parity coverage for lazy command imports plus bundled CLI smokes for fast paths and representative lazy commands.

Measurement

• Stable split-current cheap CLI p95 accepted in the optimization loop: 51.5 ms, down from prior cheap CLI p95 of roughly 290.8 to 293.5 ms.
• Post-review stable benchmark smoke: aggregate p95 312.5 ms, variance ratio 0.0290, status stable.
• Tool registry attribution showed static import/bundle evaluation dominates; deeper registry import splitting is deferred to a separate phase.

Validation

• git diff --check
• npm run test -- tests/cli-tools-latency-baseline.test.ts tests/cli-help-parity.test.ts
• npm run build
• npm run lint
• npm run typecheck
• npm run test:release-gate
• npm run test
• Real-world dist CLI smokes for help/version text/json/stream-json, quiet help/version, native status, artifacts cleanup, and structured daemon-bound status/macro-resolve errors
• RepoPrompt rereview: no remaining blocking findings or material PR-readiness risks