Protect Test Files From Unintended AI Edits

[yeus]

Problem

Our tests are spread across the repo (Rust unit tests, integration tests,
wasm/vscode smoke tests).
When AI agents make changes, it is hard to notice test modifications in
review.
Tests are a key safety net, so accidental test edits are risky.

Goal

Add simple protections so test changes are always deliberate and human-
reviewed.

Proposal

1. Add CI check: fail if test files changed without an explicit opt-in.
2. Add CODEOWNERS for all test paths so human review is required.
3. Document the rule in CONTRIBUTING.md:
   • “AI-generated PRs should not modify tests unless explicitly
     requested.”
   • “If tests must change, explain why in the PR description.”

Scope (test paths to protect)

• crates//tests/
• crates/**/*tests.rs
• crates//src//tests/**
• editors//tests/

Acceptance Criteria

• PRs that change protected test files fail CI by default.
• There is a clear, manual override path (label or env flag) for
  intentional test updates.
• At least one human reviewer is required for test-file changes.
• Contributor docs explain the policy clearly.

Why this helps

• Reduces accidental weakening or drifting of tests.
• Keeps reviews focused on intentional behavior changes.
• Works with current Rust-style distributed test layout (no major test
  reorganization needed).