Veto a one-case letter in the strict suffix test as its capital (#727)

[derek73]

Closes #727.

rules/T4 (#723) tags a bare lowercase letter in a one-case name an initial, but the strict suffix test (_vocab.is_suffix_strict) vetoes an initial by the word's text, not by that tag. So a lowercase v or i passed where its capital was vetoed. Two callers reached it:

input	policy	before	after (= its capitals)
john smith, v	lenient_comma_suffixes=False	suffix v	given v, family john smith
smith, john, i	lenient_comma_suffixes=False	no report	comma-structure
田中さん v	default	family 田中, suffix さん v	given 田中さん, family v (1.4.0's reading)

The third is W2's honorific peel, whose scan-back asks the same strict test. A CJK name with a Latin letter beside it is one case, the letter being its only cased word.

Change. is_suffix_strict takes the one-case fact and vetoes a bare lowercase letter whose capital is A to Z where it holds. is_wholly_suffix and name_word_count take it as letters_one_case, a keyword kept apart from is_wholly_suffix's one_case, which is the credential lean and is None by design at callers that want no lean. segment, _comma and script_segment pass the fact they already hold. segment asks for it only under the strict knob, and the honorific peel only for a single lowercase letter. The lenient default never reads it.

Cost. Frames are unchanged on the default path for the reference name and the common comma names (counted by the code review). 田中さん v falls from 291 to 229 frames, since the peel no longer runs. Under the opt-in strict knob, a three-part comma name pays for the case fact (smith, john, v 206 → 257); no guard covers that knob.

Tests. The one-case invariant's grid gains three comma shapes and 田中さん {n}, and runs its comma shapes again under the strict knob, which reaches nothing else (the docs review measured 0 differences on the comma-free shapes). 2444f84 disagreed on 72 of 4,554 pairs (54 strict comma, 18 honorific), this branch on none. The new control records that 72. Three case rows are added. No corpus name moved: the snapshot gains only 田中さん v.

Docs. T4 says a suffix test that vetoes an initial vetoes a one-case letter too (C1's strict mode, W2's scan-back). C1's strict-mode sentence and a strict example, decisions/T4's #727 entry (and its first entry's Open line closed), the knob's docstring, customize.rst, and the #723 release bullet extended.

Reviews. The code review found no defects. The design-docs review found the stale Open line and a docstring quoting C1's old wording; both are fixed in 5d552c1. A gap they found that predates this PR, john smith nee x x against JOHN SMITH NEE X X in the maiden take, is filed as #731.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.04%. Comparing base (2444f84) to head (5d552c1).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #729   +/-   ##
=======================================
  Coverage   99.04%   99.04%           
=======================================
  Files          46       46           
  Lines        4586     4588    +2     
=======================================
+ Hits         4542     4544    +2     
  Misses         44       44           

☔ 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.