AI tool policy: add AGENTS.md

We have the AI policy in contributing.rst and the disclosure checkbox in the PR template, but no file that AI agents actually read.

Most coding agents pick up a root-level AGENTS.md automatically. So, the proposal is to add a minimal one that only points to existing docs (AI policy + disclosure, CODING_STYLE.md, ReviewChecklist.md, how to build and run the tests, Changelog expectations), plus a CLAUDE.md symlink. We would start with a minimal one, and we can improve it with coding style and stuff like that in future.

LLVM did the same recently:

• https://discourse.llvm.org/t/rfc-llvm-ai-tool-policy-agents-md/91712
• https://llvm.org/docs/AIToolPolicy.html (the "Handling violations" part could also go into contributing.rst)