Harden runtime validation and add dashboard shutdown - #32

Problem

The portable Node repair could execute an existing unowned runtime during plan-only readiness checks, accept npm after only a version probe, and print unsafe or unusable PowerShell fallback guidance. Users also lacked a packaged way to stop a foreground dashboard from another terminal without manually identifying a Node process.

Change

• Check the FixLab ownership marker before executing any existing portable runtime.
• Require npm.cmd and a real npm pack --dry-run --ignore-scripts --json probe for existing, extracted, and final runtimes.
• Generate PowerShell-safe single-quoted fallback paths and require explicit exact-version guidance for unpinned repositories.
• Add fixlab dashboard [repository] --stop using an instance-specific local token and graceful loopback shutdown endpoint.
• Document the strengthened runtime and dashboard ownership contracts.

Validation

• Focused tests pass
• npm pack --dry-run contains the required files
• Documentation is updated
• No credentials, tokens, private endpoints, or test data are included
• npm run validate passes (92 unit tests and 8 Playwright tests)
• Existing FMDM portable runtime passes the real npm operation probe after verified ownership migration
• An unowned runtime directory is rejected before readiness inspection
• Dashboard stop E2E leaves an unrelated dashboard process running
• Fresh GitHub clone installed into an isolated prefix, started and stopped the dashboard, then uninstalled with the command removed

Risk

Low. Runtime changes add pre-install safety checks only. Dashboard shutdown accepts only a loopback request carrying the instance-specific token written by that dashboard. It never kills processes by name or PID and does not change permanent PATH, NVM, or repository content.