DRES: Benchmarking LLMs for Disfluency Removal