UPSTREAM PR #17947: scripts: add script to compare logits of llama.cpp against other frameworks

[loci-dev]

Mirrored from ggml-org/llama.cpp#17947

Compare logits between llama.cpp and another inference engine.

Example usage:
    Step 1: Dump logits from two different endpoints
        python scripts/compare-logits.py dump logits_llama.log http://localhost:8080/completions
        python scripts/compare-logits.py dump logits_other.log http://other-engine:8000/completions

        (optionally, you can add --api-key <key> if the endpoint requires authentication)

    Step 2: Compare the dumped logits
        python scripts/compare-logits.py compare logits_llama.log logits_other.log report.md

[loci-review]

Explore the complete analysis inside the Version Insights