Inferring multiple helper Dafny assertions with LLMs