feat(app): persist crash-loop state and guard app starts for safe mode (#122)

[rowkav09]

What changed

• New src/startup-recovery-store.js:
  • createStartupRecoveryStore({ file }) loads and saves the feat: add startup recovery state core #181 recovery state as a small owner-only JSON file, written atomically. A missing, corrupt or oversized file loads as a clean start, so recovery can't be what stops the app from starting.
  • guardStartup({ store, start }) counts the attempt as a failure before calling start, so a hard crash (process killed, native fault) still counts. Once the app has stayed up for healthyAfterMs (60 s by default) the count is cleared. After three unconfirmed starts in a row, start gets { safeMode: true }. A thrown startup error records which subsystem failed (configuration, provider, discord, updater or unknown, based on startupCode) and is rethrown.
  • Safe mode doesn't clear itself: a safe-mode run staying up says nothing about the parts it turned off. retryNormal() clears it, which is what the tray's "retry normal startup" will call.
  • If the state can't be saved (read-only disk), startup carries on.
• Tests cover all of the above with a manual timer, so there are no real waits.

Why

First wiring slice for #122. The follow-up PR wires guardStartup into nowplaying.exe start, has startAppFromConfig honour safeMode (no polling, Discord or updates), and adds the tray actions.

Checks

• Tests pass locally (full suite green)
• No secrets, tokens, server URLs, or personal media data are committed
• Docs or tests were updated when behavior changed
• The change is scoped to one roadmap issue

Issue

Part of #122

[github-actions]

/mira pause

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!