Improve Frontend Message Rendering and ClassificationPublic

AI Summary

The user wanted to improve how special message types like command tags, attachments, and sub-agent messages are rendered in the frontend, ensuring consistent handling and proper display in different views. The outcome includes a new centralized classification and rendering layer that hides command tags in highlights but shows them in full trace, aligns attachment spacing, updates icons for sub-agents, adds filtering options, and updates documentation. The changes were finalized, tested, and a pull request was created and reviewed.