Add Copilot APM skill discovery E2E test and harness assertions

[Copilot]

This adds coverage for Copilot APM skill discovery by validating that a skill restored under .agents/skills is discoverable and can drive workflow output. The new test closes the gap introduced by workspace-based Copilot engine launch behavior.

• New Copilot workflow test (workflow_dispatch)

  • Added test-copilot-apm-skill-discovery.md using safe-outputs.create-issue.
  • Prompt requires invoking apm-skill-discovery from .agents/skills and emitting a deterministic token in the issue body.

• APM skill fixture

  • Added .agents/skills/apm-skill-discovery/instructions.md with frontmatter + fixed token response to make discovery/invocation unambiguous.

• E2E harness integration

  • Registered test-copilot-apm-skill-discovery in e2e.sh test inventory.
  • Routed it through workflow-dispatch execution path.
  • Added body-content assertion helper and wired validation to require the expected token, not just issue creation.

• Compiled workflow + coverage docs

  • Added generated lockfile test-copilot-apm-skill-discovery.lock.yml.
  • Updated README safe-output coverage matrix to include APM skill discovery.

Use the `apm-skill-discovery` skill from `.agents/skills`.
Invoke that skill and copy its exact token response verbatim as the first line of the issue body.