Add optional argument original_output to vLLM's generate function

[LouSalaun]

This pull request fixes #1199. An optional argument original_output is added to the vLLM's generate function:

outlines/outlines/models/vllm.py

Lines 44 to 53 in a2fd35c

	def generate(
	self,
	prompts: Union[str, List[str]],
	generation_parameters: GenerationParameters,
	logits_processor,
	sampling_parameters: SamplingParameters,
	*,
	sampling_params: Optional["SamplingParams"] = None,
	use_tqdm: bool = True,
	):

original_output is a boolean. If True then the function returns the original (full) output of the vLLM API call.

[cpfiffer]

This part won't pass coverage checks because it isn't tested. Could you add a test here: https://github.com/dottxt-ai/outlines/blob/main/tests/generate/test_integration_vllm.py

[LouSalaun]

Okay just pushed the integration tests

[hedrergudene]

Any updates on this thread? It would be a highly desirable feature