[Doc] Update Batch-level DP docs

[DarkLight1337]

Purpose

• Mention that TTFT is improved as well (to avoid confusion like [Usage]: vision encoder speedup? #25746 (comment))
• Add more supported models

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 updates the documentation for batch-level Data Parallelism for multi-modal encoders. The changes include mentioning the improvement in Time to First Token (TTFT) and adding more models to the list of supported models. The documentation is now more accurate and up-to-date. However, I've noticed a potential error in the list of supported models where Llama4 is included. Llama4 appears to be a text-only model and may not be applicable for a feature targeting multi-modal encoders. Please review this point.