Credit ONNX Runtime for reranking in the acknowledgments

[felladrin]

Description

Credits ONNX Runtime for reranking in the acknowledgments table. Reranking moved off llama.cpp in #2235, but the table still credited it for that role.

llama.cpp keeps its row: wllama is llama.cpp compiled to WebAssembly and still powers in-browser inference, so only the reranking half of that row's role moves out.

Before:

Project	Role
wllama and llama.cpp	In-browser inference and local reranking

After:

Project	Role
ONNX Runtime	Local reranking of the search results
wllama and llama.cpp	In-browser inference

The new row sits directly after SearXNG, so the table follows the pipeline: aggregate the results, rerank them, then generate the answer.

Nothing else in the README needed changing. The architecture diagram already says Reranker / ONNX Runtime, and the prose describes reranking without naming an engine ("reranked locally by a cross-encoder model"). The llama.cpp mention further up is about pointing MiniSearch at a llama.cpp server as an OpenAI-compatible backend, which is unrelated.

Type of Change

• Bug fix
• New feature
• Documentation
• Other (refactor, build, chore)

How to test

1. npm run lint passes, including scripts/documentation-validator.cjs, which checks every link in README.md.
2. https://github.com/microsoft/onnxruntime returns HTTP 200.
3. Read the Acknowledgments table on this branch and confirm the two rows render as above.

Checklist

• npm run lint passes
• Tests pass (npm run test), with tests added where it made sense

No tests: the change is one table row in the README.

Security, performance, or breaking changes

None.