Divide 付, 代, 候, 芦 and 农 under locales.ZH (#952) - #975

Closes #952.

parser_for(locales.ZH).parse("付晓明") now gives family 付, given 晓明; through 2.8.0 it gave family 付晓明.

What changed

Five mainland surnames are added by hand at the tail of zh.py's census block, each annotated with its evidence as 萧 is: 付, 代, 候, 芦 and 农. With them come 蘆 and 農, the traditional forms the Taiwan ranking attests.

• They rank 65th, 106th, 187th, 197th and 198th by people in China's national ID registry (NCIIC 2008). That source is read only as a cross-check, since it is GPL/NC.
• They were the only unlisted surnames in that registry's top 200.
• Neither of the list's rankings reaches them: Taiwan ranks them 435th, 545th, 351st, 642nd (蘆) and 317th (農).
• The tool's generated block is unchanged. The hand rows sit before it, and tools/zh_surnames.py --write reproduces it byte for byte.

decisions/W1 (unspaced CJK division) gets the following:

• a dated entry with the measurements and their recipes;
• an amended Accepted bullet for the Japanese cost;
• wainshine/Chinese-Names-Corpus under "Sources not used as data", because it ranks distinct names taken from text, not people;
• an Open line for the third-source route.

Measured

All measured with parser_for(locales.ZH), comparing 4ebb51d with this branch, and asserting __file__ on each side.

• Mainland list: the 5,512 that moved are every name opening with one of the five: 付 3,488, 代 1,323, 芦 371, 农 213, 候 117.
• Japanese cost per character: 付 0, 代 14, 候 0, 农 0, 農 0, 芦 102, 蘆 21. ZH already divided 23.6% of that list, and 芦 adds about 0.06 points.
• tw-donors: 21 names moved, each from whole to divided: 農 7, 候 6, 蘆 5, 付 3. One of them is a committee named for the district 蘆洲, which now reads family 蘆.
• Default parser: no reading moves. The snapshot only gains this PR's nine case rows.
• Other records: the CLDR, Wikidata-native and per-pack agreement records all match unchanged. JA was checked too, with namedivider installed.

The list files hold one more name each than #952 counted (1,144,226 and 172,413 non-empty lines after the header). The counts above use those totals.

Tests

• Case rows for each surname under zh, plus the traditional 農建華 and 蘆建國.
• Japanese cost rows: 代田稔 → family 代, and 芦川誠 → family 芦.
• corpus_cjk.jsonl and snapshot.jsonl regenerated for the new rows.
• Checks run: uv run pytest, -n0 -m serial, mypy and ruff. The suite was also run with namedivider installed and without it.

Left open

The Open line for a mainland-weighted third source in tools/zh_surnames.py has no issue yet, and #952 closes with this PR. A draft for one is waiting for a decision on filing it.

🤖 Generated with Claude Code