[Frontend] Expose do_log_stats interval to env

[Csrayz]

Purpose

Sometimes, to track the changes in the waiting and running queues over time for specific load scenarios, it is necessary to reduce the time interval of calling do_log_stats in api_server. This is because excessively long time intervals can lead to unobservable metric changes during the interval.

For example, in scenarios with very long inputs but short outputs, the running queue may always appear as 1. However, theoretically, the running queue should show 2/3/... in some cases. This happens when a single request's size is much larger than the token budget. After the previous request's prefill is completed, it will still be in the running queue for decoding, and the next request will also be placed in the running queue for processing. At this point, it should be possible to see the running queue change from 1 to 2. However, due to the long interval between two do_log_stats calls, the decoding request has already finished, making it impossible to observe this phenomenon.

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

[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.

🚀

[gemini-code-assist]

Code Review

This pull request introduces a valuable feature by exposing the do_log_stats interval as an environment variable, which is very useful for debugging. The implementation is straightforward and the documentation update is helpful. I have one suggestion to improve the code quality in vllm/envs.py by refactoring the environment variable parsing logic to be more efficient and maintainable.