[Misc] Clean up MM profiling warnings

[ywang96]

Purpose

As we're deprecating V0, some of multimodal profiling warnings can be removed.

Test Plan

Test Result

---

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.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[gemini-code-assist]

Code Review

This pull request cleans up multimodal profiling warnings that are no longer relevant with the deprecation of the V0 scheduler. The changes remove warnings related to sequence length limitations that are now handled by the V1 scheduler's chunked prefill mechanism. The modifications are straightforward and align with the goal of removing obsolete code. I have reviewed the changes and found no high or critical issues.

[ywang96]

This warning will now show for the QwenVL model series by default since we modified profiling logic in #24312

Since we already return the following error message (without the server being crashed) if user actually sends a request longer than context window, I think this warning is no longer necessary and will be rather confusing in V1.

openai.BadRequestError: Error code: 400 - {'error': {'message': 'The decoder prompt (length 131072) is longer than the maximum model length of 128000. Make sure thatmax_model_lenis no smaller than the number of text tokens plus multimodal tokens. For image inputs, the number of image tokens depends on the number of images, and possibly their aspect ratios as well.', 'type': 'BadRequestError', 'param': None, 'code': 400}}