Add inference-perf vLLM recipes for Qwen3-32B (FP8/NVFP4) and Llama-3.1-70B-FP8 on G4

Add inference-perf vLLM recipes for Qwen3-32B (FP8/NVFP4) and Llama-3.1-70B-FP8 on G4 #114