feat(examples): add voice pipeline demo with LLM-TTS integration

- Add demo_full_voice_pipeline.py with multiple test modes:
  - TTS only synthesis
  - LLM to TTS streaming pipeline
  - VAD (Voice Activity Detection) test
  - Full pipeline simulation with mock Whisper
- Fix AudioBuffer handling in llm_tts.py (use .data.to_numpy())
- Add resolve_model_path() helper for directory-to-file path resolution
- Add tokenizer loading for LLM pipeline
- Update demo files to use correct load_model_from_safetensors API

Related to #128, #129

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: m96-chan