Add AI usage/automation policy

Recent issues and PRs are likely to be generated mechanically.
We may need to have clear policy on this.

My personal view on this:

• In general, we should not against a well-understood technology that can offload some of our repetitive work.
• At the same time, we should be guarded against massive amount of geneated issues and PRs that will drain maintainers' capacity to properly review.
• At the very least, I would suggest that the issue/PR summary should be written by human. So it will express genuine concern for an issue or demonstrate understanding of the proposed changes. Any use of AI tool should be declared (the use of Copilot review is explicit in itself, as GitHub interface will show it as reviewer or in some case a co-author).

Start by reviewing this issue together with the recent issues and PRs it references. No file, test, or documentation entry point is named; the work is complete only after maintainers agree on the policy and publish it in the appropriate project documentation.