Add FP8 dynamically quantized Qwen3-0.6B-Base model using llm-compressor

initial commit