Draft: chore(skills): add e2e-tests skill for catalog-paired end-to-end tests

Add the e2e-tests skill: generate and maintain Artifact Registry end-to-end tests from the docs/testing/e2e/ scenario catalogs, keeping the catalog (the UAT contract) and the test suite in sync via the scenario ID as the join key. Bundles audit_catalog_sync.py (the drift check) and references for the harness layout, Ginkgo, and Playwright conventions.

Summary

This merge request introduces a new skill definition for generating and maintaining end-to-end (e2e) tests for the Artifact Registry. It establishes clear rules for keeping test code in sync with a "scenario catalog" — a set of documented user journeys that serve as the official contract for what the product should do.

The core idea is simple: every scenario in the catalog that can be automated must have exactly one matching test, linked by a unique ID. If a test and the catalog disagree, the catalog always wins. The skill also defines when not to write tests (e.g., when a scenario is blocked or intentionally manual), and what to do when the test infrastructure doesn't exist yet (document the intent, don't force it).

It also includes style guidelines for writing the actual tests, emphasizing that tests should verify what the user experiences, not low-level technical details.