With the Realtime With WebRTC API, The event response transcription is sometimes recieved before the event input transcription

With the realtime api ```
gpt-4o-mini-realtime-preview-2024-12-17


Example:
```json
{
    "type": "response.audio_transcript.done",
    "event_id": "event_B5sUl44TeyVrwqsUYPR6x",
    "response_id": "resp_B5sUj5CfptJftECDjlJoj",
    "item_id": "item_B5sUjPvx0oz9H9FNPYOCR",
    "output_index": 0,
    "content_index": 0,
    "transcript": "Quoi encore ? Encore un problème de qualité ? Écoute, je n'ai pas le temps de m'assouplir. On doit régler ça rapidement. Dis-moi ce qui ne va pas et on va trouver une solution. Allez, fais vite."
}

{
    "type": "conversation.item.input_audio_transcription.completed",
    "event_id": "event_B5sUlwikJFM5dbomnhcb2",
    "item_id": "item_B5sUiFBCF29tqfeET1UPL",
    "content_index": 0,
    "transcript": "Bonjour\n"
}

The event “response.audio_transcript.done” is recieved before the event “conversation.item.input_audio_transcription.completed”.