Read halfwidth corner brackets as a nickname delimiter (#597)

[derek73]

The full-width corner brackets 「」 have been a default nickname pair since #273, but their halfwidth forms ｢｣ (U+FF62/U+FF63) were not. Those halfwidth forms are the only corner brackets legacy JIS X 0201 data has. So a clause enclosed in them stayed in the name, brackets and all.

Fix. DEFAULT_NICKNAME_DELIMITERS gains ("｢", "｣"), and the v1 facade's nickname_delimiters gains the key halfwidth_corner_brackets, so HumanName reads the pair the same way parse() does.

Input	2.3.0	This PR
山田 ｢ﾀﾛｰ｣ ﾀﾛｳ	given 山田, middle ｢ﾀﾛｰ｣	family 山田, given ﾀﾛｳ, nickname ﾀﾛｰ
山田｢ﾀﾛｰ｣太郎	one given word	family 山田, given 太郎, nickname ﾀﾛｰ
John ｢Jack｣ Smith	middle ｢Jack｣	nickname Jack

The order change in the first row comes from #594 and this PR together. #594 (halfwidth katakana) is unreleased, in the same cycle. Released versions read that name given-first because the halfwidth kana were unclassified. After #594 alone, the unclassified bracket was the only thing still stopping the name from reading as Japanese.

What's not covered:

• A v1 Constants restored from a pickle keeps the keys it was saved with, so it doesn't gain the new pair. That's the same policy Support smart quotes (“Jack”), guillemets («Petit»), and CJK brackets as nickname delimiters #273's pairs followed.
• Unicode has no halfwidth form of 『』 or of the fullwidth parentheses, so nothing else needs a halfwidth twin.

Verification

• Three fix(#597) case rows, run through both parse() and HumanName. With the facade key removed, the three facade rows fail, and so does test_snapshot_field_translation.
• The reviewer checked that halfwidth and full-width brackets read identically: every corpus and case text containing 「」, re-spelled with ｢｣, gives the same output in both APIs. No differences across 9 texts.
• fix(#597) ledger rules at all five baselines. No corpus line held the brackets before this, so the three new rows are the only names that move. The gate exits 0 everywhere.
• 11,180 tests pass, plus mypy, ruff and the Sphinx doctests.
• Docs: rules.md N1 gets an example and a history: link, plus a new decisions.md#N1 entry. modules.rst, AGENTS.md's note on the v1 delimiter keys, and the release log are updated too.
• Commit 2 fixes what the pre-PR review found. All of it was prose: who caused the order change, "kept as a middle" being true only for the spaced rows, the pickle caveat, and two shim comments.

Closes #597.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.98%. Comparing base (8071645) to head (e1447f2).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #599   +/-   ##
=======================================
  Coverage   98.98%   98.98%           
=======================================
  Files          45       45           
  Lines        4235     4235           
=======================================
  Hits         4192     4192           
  Misses         43       43           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.