Clean up dead Qwen local model fallback and add .gitignore

Remove llama-cpp-python and disable local Qwen fallback parsing to fix C++ compilation time on HF Spaces

Deploy Local Hybrid Engine