[MISC] Consolidate cleanup() and refactor offline_inference_with_prefix.py

[comaniac]

When refactoring offline_inference_with_prefix.py due to #9505, I found that cleanup() has been implemented many times in contests. This PR consolidates them to a utility function.

cc @tlrmchlsmth @DarkLight1337 @youkaichao

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[tlrmchlsmth]

should be able to hit test_lazy_outlines.py with a cleanup as well:

vllm/tests/entrypoints/llm/test_lazy_outlines.py

Lines 17 to 34 in ae8b633

	llm = LLM(model="facebook/opt-125m",
	enforce_eager=True,
	gpu_memory_utilization=0.3)
	outputs = llm.generate(prompts, sampling_params)
	for output in outputs:
	prompt = output.prompt
	generated_text = output.outputs[0].text
	print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
	# make sure outlines is not imported
	assert 'outlines' not in sys.modules
	# The second LLM needs to request a higher gpu_memory_utilization because
	# the first LLM has already allocated a full 30% of the gpu memory.
	llm = LLM(model="facebook/opt-125m",
	enforce_eager=True,
	guided_decoding_backend="lm-format-enforcer",
	gpu_memory_utilization=0.6)

[tlrmchlsmth]

Thanks for the nice cleanup!

[comaniac]

CI failed due to #9424 which is unrelated