fix(stt): set scribe_v2_realtime model with language_code for streaming STT

- Add model_id="scribe_v2_realtime" (already set) + language_code from STT_LANGUAGE env (default "de")
- Remove _stt_session from cleanup loop (plugin uses livekit http_context)
- Remove _stt_session stub from __init__

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

voice.py

@@ -231,7 +231,6 @@ class VoiceSession:
         self.session = None
         self._task = None
         self._http_session = None
-        self._stt_session = None  # kept for cleanup compat
         self._caller_key: bytes | None = None
         self._caller_identity: str | None = None
         self._caller_all_keys: dict = {}  # {index: bytes} — all caller keys by index
@@ -330,7 +329,7 @@ class VoiceSession:
         self._task = asyncio.create_task(self._run())
 
     async def stop(self):
-        for obj in [self.session, self.lk_room, self._http_session, self._stt_session]:
+        for obj in [self.session, self.lk_room, self._http_session]:
             if obj:
                 try:
                     if hasattr(obj, "aclose"):
@@ -557,7 +556,11 @@ class VoiceSession:
             self._http_session = aiohttp.ClientSession()
             voice_id = os.environ.get("ELEVENLABS_VOICE_ID", DEFAULT_VOICE_ID)
             self.session = AgentSession(
-                stt=elevenlabs.STT(api_key=ELEVENLABS_KEY, model_id="scribe_v2_realtime"),
+                stt=elevenlabs.STT(
+                    api_key=ELEVENLABS_KEY,
+                    model_id="scribe_v2_realtime",
+                    language_code=os.environ.get("STT_LANGUAGE", "de"),
+                ),
                 llm=lk_openai.LLM(base_url=LITELLM_URL, api_key=LITELLM_KEY, model=self.model),
                 tts=elevenlabs.TTS(voice_id=voice_id, model="eleven_multilingual_v2",
                                    api_key=ELEVENLABS_KEY, http_session=self._http_session),