fix: add STT timeout, move finally restart to thread, guard exit on recording

- Set OpenAI client timeout=30s in transcribe_audio() — default 600s
  blocks _voice_processing for 10 min if Groq/OpenAI stalls
- Move _voice_start_recording in _voice_stop_and_transcribe finally
  block to a daemon thread (same pattern as Ctrl+B handler and
  process_loop)
- Add _should_exit guard at top of _voice_start_recording so all 4
  call sites respect shutdown without individual checks

tools/transcription_tools.py

@@ -150,7 +150,7 @@ def transcribe_audio(file_path: str, model: Optional[str] = None) -> Dict[str, A
     try:
         from openai import OpenAI, APIError, APIConnectionError, APITimeoutError
 
-        client = OpenAI(api_key=api_key, base_url=base_url)
+        client = OpenAI(api_key=api_key, base_url=base_url, timeout=30)
 
         with open(file_path, "rb") as audio_file:
             transcription = client.audio.transcriptions.create(