refactor: remove hosted benchmark and scenario CLI commands

[wall-rl]

Description

Remove benchmark jobs and scenario commands, interactive screens, stores, services and URL helpers. Remove their menu entry and shortcut, product-only tests and scoring status labels. Regenerate command documentation and test the surviving command registry.

Delete the checked-in generated MCP bundle, which embedded a stale SDK, and ignore it like the extension archive. The existing build:mcp command regenerates the bundle before packaging; the MCP source and supported devbox tools remain intact.

Motivation

The hosted benchmark product is being removed outright. API dependency: https://github.com/runloopai/runloop/pull/11947; documentation companion: https://github.com/runloopai/docs/pull/259. Coordinate the next CLI release with the breaking SDK/API removal. The pinned SDK dependency should be refreshed through its normal generated release, not by editing bundled SDK internals.

Testing

• TypeScript build and MCP bundle/archive build pass with the pinned pnpm version.
• 68 hermetic Jest suites: 735 tests pass. The aggregate runner reported an open-handle warning afterward; the new command registry test passes and exits with --detectOpenHandles.
• ESLint: no errors (72 existing warnings across the repository).
• Changed files pass Prettier; git diff --check passes.
• The default test command also includes a live scp/rsync suite; that initial run failed without its external test infrastructure. Subsequent validation excluded /e2e/; no live test is needed for this change.

Complete removal stack

The full five-PR backend stack and eight linked companion PRs are tracked in NUKE_BENCHMARK.md. Cross-repository PRs are rollout dependencies, not GitHub base branches. No PR applies live infrastructure destruction or publishes SDK packages.

[reflex-loop]

🔄 Reflex agent status: Working

Agent	Status
agt_5wS1r77ya5tLkapia60opU	🔄 Working

This PR was created by Reflex.

_{This comment updates in place as agents work.}