Remove scipy dep by implementing resample_poly

[mgoin]

The only reason why scipy is a common dependency of vLLM is because phi4mm requires it for resample_poly (see discussion in the PR #14119 (comment)). Since it isn't that complex of a function, we can implement it in a couple dozen lines of numpy and remove scipy.

Needs more testing to validate

[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.

🚀

[Isotr0py]

I'm afraid that if we finally refactor Phi-4-MM to use merged processor in #15477, scipy will still be a necessary dep to make hf_processor work, because it's at the import level in processing_phi4mm.py.
(https://huggingface.co/microsoft/Phi-4-multimodal-instruct/blob/main/processing_phi4mm.py#L24)

[Isotr0py]

@mgoin FYI, HF plans to convert Phi4-mm model repo to HF format, which includes fast image processor and feature extractor implemented with torch, so that we no longer need scipy in requirements after that. (https://huggingface.co/microsoft/Phi-4-multimodal-instruct/discussions/70)

I'm starting to implement HF format Phi-4-MM here: #17121, so let's leave scipy to be removed after this migration!