fix: gpt-realtime-whisper requires Realtime API, not REST endpoint

Swapped to gpt-4o-transcribe (/usr/bin/bash.006/min) — middle ground between
speed and cost.

backend/main.py

@@ -85,7 +85,7 @@ async def transcribe_audio(audio_bytes: bytes) -> str | None:
     client = get_openai()
     try:
         transcript = await client.audio.transcriptions.create(
-            model="gpt-realtime-whisper",
+            model="gpt-4o-transcribe",
             file=("audio.webm", audio_bytes, "audio/webm"),
             response_format="text",
         )