[MISC][V1] Handle exception of current_platform.get_device_name() in arg_utils

[comaniac]

The use of current_platform.get_device_name() in arg_utils.py is too early and is not compatible with Ray, because Ray Serve may import vLLM at the main actor to use some of its data structures. Since this actor doesn't have GPU and is not the actor that actually runs the engine, we will get an exception about no GPU devices. This PR makes this early device query optional to workaround this issue. It should not affect any existing use cases and should not impact user experience, given that the only difference without querying device in EngineArgs._override_v1_engine_args is default_max_num_batched_tokens.

cc @WoosukKwon

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

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 either: Add ready label to the PR or enable auto-merge.

🚀

[njhill]

cc @youkaichao