Restructure PaperReading into an evidence-grounded research workflow

[AOROM]

What changed

• repositioned PaperReading from a Codex-only skill into an evidence-grounded research workflow for empirical social science
• added a reusable, dependency-free paperreading Python core with PaperRecord, EmpiricalDesign, Finding, and EvidenceRef
• added evidence labels and evidence-coverage metrics
• added a cross-paper Literature Matrix builder and Markdown renderer
• added schemas/paper.schema.json and a sanitized structured example
• added ROADMAP.md, CHANGELOG.md, CITATION.cff, MIT LICENSE, architecture documentation, and a feature-request issue template
• rewrote the English and Chinese READMEs around the new product positioning while preserving the existing Codex skill and Excel workflow
• extended CI compilation and added core tests

Why

The original repository already had a strong 13-field empirical-paper reading contract and safe Excel writer, but its value was framed too narrowly around a single Codex skill. This restructuring makes the 13-field workflow one adapter/export path inside a broader, portable research-intelligence architecture, creating a foundation for Evidence Maps, method auditing, batch comparison, research-gap synthesis, and future integrations.

Compatibility

• existing skills/papers-reading-skill/ behavior is unchanged
• existing workbook-writing behavior is unchanged
• the new Python core uses only the standard library

Validation

CI is configured to run:

• python tools/validate_skill.py skills/papers-reading-skill
• python -m compileall -q paperreading skills tools tests
• python -m unittest discover -s tests -v

This PR is intentionally draft while CI and the new public-facing architecture are reviewed.