Fix reasoning-only CodeLayer responses

[K-Mistele]

Summary

• Continue AgentLayer runs after reasoning-only assistant responses instead of treating them as complete.
• Increase default Codex GPT-5.5 reasoning effort to medium for more reliable reasoning summaries.
• Extend mock-provider coverage for reasoning-only responses followed by final text.

Validation

• bun --bun test packages/agentlayer-core/test/streaming-events.test.ts agents/codelayer/test/agent.test.ts
• bun --bun run typecheck
• pre-push hook: bun run check

[github-actions]

Docs Agent Review

I'll examine the existing documentation to see if it covers these changes. Let me start by exploring the docs structure.

---

To apply these recommendations, comment: @docs-agent apply