[Bugfix] Fix various bugs in multi-modal processor

[DarkLight1337]

Fixes:

• Placeholder search results in duplicate outputs when different modalities share the same placeholder tokens
• _apply_prompt_replacements being called unconditionally because of iterating over .items() instead of .values()
• Unable to call get_max_tokens_per_item_by_modality for models that require trust_remote_code=True

Thanks @HwwwwwwwH for finding these bugs!

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
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 do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[Isotr0py]

LGTM!