[Web][Backend] Voice command accessibility mode -> Web Speech API + preference endpoint (closes #287, #286)

[mustafa-5493]

📝 Description

Closes #287 and parent #286. Implements requirement 1.1.3.8 end-to-end:

Backend

• voiceCommandsEnabled boolean column on RegisteredUser (default false) + V6 Flyway migration
• VoiceCommandsRequest DTO + RegisteredUserService.setVoiceCommandsEnabled
• Owner-only PATCH /api/users/me/voice-commands (mirrors the leaderboard-visibility pattern)
• voiceCommandsEnabled surfaced in UserProfileDTO so the web client reads it via /me

Web

• parseVoiceCommand -> pure deterministic parser: zoom in/out, my location, find/search <q>, navigate to/go to <X>, report (<cat>) here
• useSpeechRecognition -> thin React wrapper over window.SpeechRecognition / webkitSpeechRecognition, auto re-arms on quiet pauses, cleans up on unmount, surfaces errors as state
• VoiceCommandContext -> app-level { active, lastEvent } state bridging the navbar mic toggle and the per-page command runner
• VoiceCommandRunner -> lives inside the Leaflet MapContainer; dispatches recognized commands to existing Home setters (map.zoomIn(), setSearchTarget, setRouteDest, setShowCreatePanel, etc.) so a voice command is exactly equivalent to the click path
• VoiceCommandMicButton -> navbar pill; hidden when feature unsupported or user hasn't opted in. Animates while listening (respects prefers-reduced-motion via motion-safe:)
• VoiceCommandLiveRegion -> visually-hidden polite ARIA region; announces "Heard: …" / "Zoom in" / "Command not recognized: …" / errors
• Accessibility section on /profile with a toggle calling useSetVoiceCommandsEnabled; unsupported-browser notice rendered inline when window.SpeechRecognition is missing

📊 Type

New feature

✅ Acceptance Criteria

• Voice mode can be enabled/disabled from /profile; setting persists to backend and restores on next login from /api/users/me
• Mic toggle button only shows when user is opted in and the browser supports the API
• Voice commands trigger search, navigation, report, zoom, locate flows
• Listening indicator (animated mic) visible while recording
• ARIA live region announces recognition results
• Graceful fallback message shown on unsupported browsers (Firefox/Safari without polyfill)
• All voice-triggered actions remain accessible via keyboard/mouse (no fallback regression)
• 481 frontend tests pass (15 new), backend tests cover service + controller paths

🧪 How to Test

Automated

cd backend && ./mvnw test            # service + controller tests for the new endpoint
cd frontend && npm run test:run      # 481 tests including 15 new parser tests

Manual (Chrome or Edge required for Web Speech API)

1. Bring up the stack: docker compose -f docker-compose.local.yml --env-file backend/.env up -d --build
2. Sign in, go to /profile → Accessibility section → enable "Voice commands"
3. Reload -> toggle stays on (backend persisted)
4. Mic icon appears in navbar. Click → grant mic permission → indicator pulses
5. Say each:
   • "Zoom in" / "Zoom out" → map zooms
   • "My location" → map flies to user
   • "Find pharmacy" → map flies to first geocode result
   • "Navigate to Taksim" → route panel opens with Taksim as destination
   • "Report obstacle here" → create-report panel opens
   • Gibberish → live region announces "Command not recognized"
6. Switch to Firefox → toggle in /profile shows unsupported-browser notice; mic icon hidden
7. Hard-refresh / disable voice mode → mic icon disappears, listening session ends

📸 Screenshots

⏱️ Estimated Review Time

15 min (commit-by-commit on the Commits tab -> 12 atomic commits cover backend column → DTO → service → endpoint → tests → parser → hook → context → service-call → mic UI → runner → settings toggle)