Fix voice input for iOS Safari and prevent infinite loop

- Remove temperature parameter from GPT-5-mini activity extraction (not supported)
- Add classification state to useVoiceInput hook to avoid duplicate API calls
- Prevent infinite loop in VoiceFloatingButton by tracking lastClassifiedTranscript
- Use classification from backend directly instead of making second request
- iOS Safari now successfully transcribes with Azure Whisper and classifies with GPT-5-mini

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

maternal-app/maternal-app-backend/src/modules/voice/voice.service.ts

@@ -174,7 +174,6 @@ If the text doesn't describe a trackable activity, respond with:
           { role: 'system', content: systemPrompt },
           { role: 'user', content: userPrompt },
         ],
-        temperature: 0.3,
         response_format: { type: 'json_object' },
       });