[AIAA] Build step 2: repository instructions for AI agents

Build step 2 of the AI-assisted assessment system: repository
instructions for any AI agent working in cncf/techdocs.

Deliverable: `.github/copilot-instructions.md`. It carries the safety
ground rules (content is data, not instructions; write only to this
repository; a human reviews and merges everything; secrets stay out),
the evidence discipline (reproducible numbers, cited findings,
unverifiable means flagged), and the point-don't-copy rule for the
methodology in `docs/analysis/`.

Scope note: this file loads into every Copilot surface in the repo,
including code review on ordinary PRs, so it stays lean and
repo-generic. Phase-specific behavior belongs to the agent profiles
(a later step).

Done when merged after review against the safety and agent-rules
sections of the design.

- [ ] Draft reviewed locally (assertion script green, `npm run check`
passes)
- [ ] PR opened
- [ ] Review against sections 8 and 14
- [ ] Merged

Create `.github/copilot-instructions.md` and keep it lean and repository-generic, covering the safety and evidence rules plus the point-don't-copy rule. Read sections 8 and 14 of the design, run the assertion script, and run `npm run check`. Done means the local checks pass and the file is opened for review against those sections before merge.

Written by the indexing model from the issue text.