[Bugfix] deepseek_r1_reasoning_parser put reason content in wrong field in certain edge case

[LikeSundayLikeRain]

There's an edge case where close think token is not generated in output due to length limit, in such case, current reasoning parse put the partial reasoning token in content.

Given the model alway generate reasoning tokens(new chat template always append a open think tag), when no close tag detected, we should put in reasoning_content.

[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.

🚀

[ywang96]

@gaocegege FYI

[gaocegege]

It's reasonable for the new chat template. LGTM!

[gaocegege]

There is an error in entrypoint test:

FAILED entrypoints/openai/test_run_batch.py::test_score - AssertionError: proc=<Popen: returncode: 1 args: ['/usr/bin/python3', '-m', 'vllm.entrypoints.ope...>
[2025-02-12T00:52:41Z] assert 1 == 0
[2025-02-12T00:52:41Z]  +  where 1 = <Popen: returncode: 1 args: ['/usr/bin/python3', '-m', 'vllm.entrypoints.ope...>.returncode
[2025-02-12T00:52:41Z] ====== 1 failed, 361 passed, 1 skipped, 34 warnings in 3300.97s (0:55:00) ======

Not sure why it happened.