feat(tts): switch to flash model + speed 1.15x for snappier voice

- Model: eleven_multilingual_v2 → eleven_flash_v2_5 (lower latency)
- Speed: 1.15x via VoiceSettings
- Stability/similarity tuned for natural German speech

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

voice.py

@@ -588,9 +588,12 @@ class VoiceSession:
                 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",
+                    model="eleven_flash_v2_5",
                     language="de",
                     api_key=ELEVENLABS_KEY,
+                    voice_settings=elevenlabs.VoiceSettings(
+                        stability=0.5, similarity_boost=0.75, speed=1.15,
+                    ),
                     http_session=self._http_session,
                 ),
                 vad=_get_vad(),