fix: recover stalled Cloud connections and report live status

[renyuanz]

A Cloud connector could remain alive while repeatedly dialing stale edge addresses, leaving encrypted pairing unavailable until a manual restart. The local service now supervises the connector, probes the pinned encrypted route, and replaces it after three consecutive failures when the local Hub remains reachable. Restarts back off from 15 seconds to 10 minutes and reset after five minutes of healthy checks; shutdown kills and reaps the child, and the registration lock remains inherited by the connector.

Cloud status now requires a fresh encrypted identity and route check before claiming verification. The desktop panel refreshes visible status and clears readiness after a failed refresh. Previously verified addresses and pairings are retained for reconnection. No private Cloud service code or credentials are included.

Prepares App 0.7.4 and Hub/CLI 0.21.4, including the mobile navigation already merged in #475. README, website and troubleshooting instructions describe the resulting behavior.

Validation:

• Hub tests: 175 passed plus the CLI encrypted-preflight integration test; the optional official-cloudflared ingress validation also passed.
• App TypeScript check and 21 installer/download routing tests passed.
• Container Chromium regression: 8 passed, including Cloud setup, loss/recovery of readiness, and narrow/wide mobile navigation.
• Process tests cover sustained outages, replacement/reaping, child exit, interrupted probes, stopped origin and retry backoff.