Openautotag

Create a scaffold for a horizontally scalable PDF accessibility tagging pipeline. I need this system to be contract-first, allowing Codex agents to implement modules in parallel without touching shared runtime code. The architecture should feature isolated CLI boundaries for modules and an orchestrator that runs them as separate processes. Key features to include are the baseline implementation for parser, layout analysis, semantic mapping, reading order, tag tree construction, and font embedding planning. Specifically, I need the new `font-embedder` stage, which analyzes fonts and emits remediation plans for PDF/UA compliance. Also, ensure the pipeline records stage timelines, handles retries, and emits structured diagnostics, including a `sourceTextMap` diagnostic.