Label the CJK honorific rows for the issues they depend on

[derek73]

Closes #324.

29 rows carried fix(#307) or fix(#308) alone while the ledger describes the same shapes as compounding with #271, and cases.py has had compound notation since #273 for exactly this pattern — one issue enables, #271 orders.

Which rows, measured

The issue warned that a blind sweep would fix one row into a different inconsistency, which is why #320 left it alone. So the dependency is measured, not assumed: re-parse each row with script_orders and segment_scripts removed — the documented way to restore 2.0's reading — then remove the HAN entry, the HANGUL entry and hangul segmentation one at a time.

rows whose asserted fields change	26 → compound with fix(#271)
rows unaffected	3 → left single-issue
rows depending on the HIRAGANA entry	0 → #272 appears nowhere

That last row is worth stating, because the Japanese cases look like #272 and are not: 田中 さん and 山田太郎様 key on HAN order — the name pieces are Han, and the kana is only the honorific being peeled.

The three exemptions are now deliberate

Each is a row whose note already explained why it differs; each now says so in the sweep's terms, so a future reader finds the reason in the row rather than re-deriving it:

• latin_stem_glued_kana_honorific (Andersonさん) and latin_stem_glued_hangul_honorific (Anderson선생님) — a Latin remainder never reaches the order table
• ko_honorific_after_comma (김민준, 씨) — the comma decides the family before segmentation would

One departure from the issue text

The issue's example string is fix(#271) + fix(#307); this uses fix(#307) + fix(#271), matching the two existing feat(#273) + fix(#271) rows. Those read causally — what creates the shape, then what orders it — and the issue's actual argument is consistency with that precedent. Easy to flip if you'd rather have #271 lead.

Scope

Labels only. classification is consumed by nothing but pytest failure messages, and nothing cross-checks it against the ledger's issue values (tools/differential/README.md calls the correspondence a convention; no test enforces it), so the sweep cannot break a guard.

Full suite 3096 passed; mypy and ruff clean; differential harness clean at 0 unexplained.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.50%. Comparing base (fe4fc8d) to head (0ec6106).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #363   +/-   ##
=======================================
  Coverage   98.50%   98.50%           
=======================================
  Files          44       44           
  Lines        2883     2883           
=======================================
  Hits         2840     2840           
  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.