feat: add notification sound profiles and speech

[leoncheng57]

Summary

• add versioned per-device sound and speech preferences with deterministic WebAudio profiles, per-event controls, and capability-aware previews
• migrate the shipped BFF browser.sound and browser.volume exactly once only when the device key is absent; present or corrupt device storage remains authoritative
• keep the existing single SSE watcher, server-backed browser/ntfy delivery gates, ntfy behavior, and notification deep links
• document that iOS constructor Notifications still require installed-PWA/service-worker support and ntfy remains the reliable phone path

Privacy

Speech is off by default and speaks only fixed generic status phrases. It never receives or speaks event payloads, prompts, paths, filenames, command arguments, tool output, repository paths, secrets, or notification bodies.

Verification

• npm run typecheck passed
• npm test: 22 files, 208 tests passed
• npm run build: production client and server builds passed
• CI=1 npm run test:e2e -- --workers=1: 126 passed, 1 skipped
• notification media E2E: 5 passed, covering absent-key migration, existing-key preservation, corrupt-key recovery, privacy-safe phrases, previews, disabled behavior, and 390px layout
• local /settings/notifications screenshot: 1 captured and validated on isolated ports
• git diff --check passed

/settings/notifications

[github-actions]

PR screenshots

Route	Preview
/settings/notifications Open full size

Source: b33dec5 | Download artifact

Captured from the production app against deterministic mock OpenCode fixtures only.