Sync runner lifecycle and display-id workflows

[joyzoursky]

Summary

• Synchronize runner stop semantics between CLI and control-plane so skytest stop runner marks the server runner offline immediately.
• Introduce runner displayId as first-class pairing data and use it consistently in Team Runners UI and CLI workflows.
• Reduce stale scheduling edge cases by requiring runner status ONLINE at claim time.

Changes

• Added new authenticated endpoint POST /api/runners/v1/shutdown to set runner status to OFFLINE and return protocol compatibility metadata.
• Extended runner protocol with shutdownRunnerRequest/Response and required displayId on pairing exchange payloads.
• Updated CLI runtime to:
• send shutdown notifications before local process termination (serverMarkedOffline surfaced in stop output)
• accept --url as alias for --control-plane-url
• resolve runner identifiers by local ID, server ID, or unique prefix across start/stop/logs/describe/unpair
• Persisted Runner.displayId in schema/backend types/services and surfaced it via team availability + MCP inventory responses.
• Updated Team Runners UI to:
• treat stale as offline presentation
• move pairing flow into guided modal with copyable commands
• add troubleshooting command references and per-offline-runner start command snippets
• Added/updated docs for new CLI flags and runner identifier semantics.
• Added targeted tests for CLI parser alias and claim SQL online guard.

Validation

• Lint: npm run lint (via npm run verify) passed
• Tests: targeted test updates included (packages/cli-runner/src/cli-parser.test.ts, src/lib/runners/__tests__/claim-service.test.ts)
• Audit: npm run audit (via npm run verify) passed
• Manual: Not run in this PR flow

Breaking Changes

• Runner.displayId is now a required database field and pairing exchange contract field.
• Deployments with existing Runner records require a migration/backfill plan before applying this schema change.

Risks

• Shutdown notify is best-effort; if control-plane notify fails, local stop still succeeds and UI falls back to heartbeat freshness windows.
• Identifier prefix matching may surface ambiguity errors for similar runner IDs and requires users to provide a longer ID.

Follow-ups

• Add integration coverage for shutdown endpoint + CLI stop path across network/error scenarios.
• Add migration/backfill task for pre-existing runner rows in non-empty environments.