Summary

Closes #663.

This updates the catalogue schema, all existing dataset records, and the website consumers for the requested attribute changes.

What changed

• adds Partial and the conversations unit
• renames the existing source-oriented Domain values to Source
• adds topical Domain values: General, Legal, News, Quran, and Science
• restricts Collection Style to either crawling or manual curation
• adds Annotation Style, including no annotation
• renames Test Split to Has Splits
• renames existing dialect-specific Subsets to Dialect Subsets
• adds a separate generic Subsets field for non-dialect dataset subsets
• normalizes existing venue names and defines them as schema options, with other as an escape hatch
• updates cards, search, statistics, fuzzy search, plots, validation, and README documentation

Migration rules

• existing source categories are preserved under Source
• Domain defaults to General, with conservative classification for clear News, Quran, Legal, and Science records
• Collection Style is crawling when crawling was previously present; otherwise it is manual curation
• existing human, machine, and LLM annotation values move to Annotation Style; entries without an explicit annotation method use no annotation, while unknown values remain other
• existing dialect subset data moves to Dialect Subsets; generic Subsets starts empty
• existing entries default to Partial: false

Validation

• USE_LOCAL_SCHEMA=true python validate_schema.py
• node --check on all modified JavaScript files
• python -m py_compile validate_schema.py plots.py
• full catalogue audit confirming all 1,120 tracked dataset files contain the new required fields and no Test Split field

Jekyll rendering was not run because Ruby/Bundler is unavailable in the local environment.