feat: package the crawler as a multi-source aroma data skill

[rastagan-git]

Summary

• Replace the four standalone workbook scripts with an installable flavor-data-crawler package and a unified flavor-data CLI.
• Preserve the original entry points as compatibility wrappers while adding typed statuses, provenance, caching, retries, atomic writes, and recovery checkpoints.
• Add source adapters for PubChem, selected pinned Pyrfume archives, and pinned M2OR receptor-assay evidence alongside the existing NIST and MFFI workflows.
• Add an English-first README with a reciprocal Chinese README, a reusable Codex skill, packaging metadata, offline tests, and cross-platform CI.

Compatibility

• The original .py and .bat entry points remain available with their established workbook defaults.
• Commands create sibling outputs and refuse accidental overwrites unless --force is explicit.
• ChemicalBook compatibility remains permission-gated because the current robots policy excludes the legacy routes; the implementation does not bypass CAPTCHA or access controls.

Data sources

• PubChem: identifiers, properties, synonyms, and attributed odor annotations.
• Pyrfume: allowlisted, commit-pinned archives with archive-level source and rights notes.
• M2OR: commit-pinned molecule–olfactory-receptor assay evidence with species, responsiveness, receptor, and DOI context.
• NIST and MFFI: retained and wrapped with conservative access behavior and provenance.

Validation

• ruff check .
• ruff format --check .
• pytest -q — 46 passed
• Official skill validator — passed
• Package dependency check and wheel build — passed
• Live smoke checks — PubChem, NIST, Pyrfume, and M2OR returned expected records

Licensing and access

No project license is added by this PR, and no downloaded upstream dataset is bundled. Provider-specific citation, access, and redistribution notes remain explicit in the documentation and output provenance.

Closes #1