[BugFix] Set default random seed to 0 for V1

[WoosukKwon]

Currently, the random seed is unset (seed=None) by default. This behavior was originally intended for vLLM V0, where the driver (rank-0) worker shared the same process as the user, causing any fixed seed to affect the user process.

In vLLM V1, however, each worker typically runs in its own process different from the user process (except for TP=1 and VLLM_ENABLE_V1_MULTIPROCESSING=0). When the seed is unset in V1, tensor parallel workers independently sample different tokens, which previously didn't matter since we only used output from the rank-0 worker and discarded others. However, this randomness causes problems for advanced features like spec decoding, where consistent token generation across TP workers is required.

This PR resolves the issue by setting the default random seed to 0 in vLLM V1.

Note: This change affects the user experience. Previously, users saw entirely different outputs for each run due to the unset seed. After this PR, first few outputs can be consistent across runs by default (while the full consistency/reproducibility requires VLLM_ENABLE_V1_MULTIPROCESSING=0).

[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.

🚀

[DarkLight1337]

LGTM now