Read a one-case bare letter as the initial its capital is (#723)

[derek73]

Closes #723.

In a name written wholly in one case, a bare capital was tagged an initial and a bare lowercase letter was not (except the roman-numeral letters, #721). So every rule that reads an initial read the two spellings of one name differently:

input	before	after (= its capitals)
j nee smith	family j	given j (M4)
j	reports given-or-family	no report (O5)
jose j y lopez	family j y lopez	middle j y, family lopez (P3's count)
smith, j y lopez	given j y lopez	given j, middle y lopez

Change. classify tags a bare letter an initial where its capital matches the bare-capital shape (A to Z), in a name whose own words are written in one case. This replaces #721's numeral-only test. A letter outside the own words (a maiden clause, a nickname, words past a declined marker) asks the whole name's case instead, so a lowercase y in MARIA GARCIA NEE LOPEZ y FERNANDEZ stays a connective. é and ж are unchanged; P3's fork still answers first for a connective letter among the own words.

What moved. Lowercase letters in names whose own words are one case, each onto its capital twin's reading. No capital spelling moved, no corpus name moved (the snapshot gains only the new rules examples), and no corpus name's capitalized() moved. A one-case clause connective now repairs as its capital does (jose lopez nee y garcia → Jose Lopez (Y Garcia)). The clause fix also returns MARY SMITH NEE v JONES to the reading 1.4.0, 2.4.2 and its mixed-case spelling give (no maiden), which only the unreleased #721 had moved. Details and measurements: docs/design/decisions/T4-bare-letter-as-an-initial.md.

Measured. A scratch grid of letters a–z, é and ж in 35 one-case shapes, three heads and three orders (recipe in decisions/T4): 504 of 7,794 lowercase/capital pairs disagreed on 16ee285, none now. The committed invariant test_a_one_case_name_reads_as_its_capitals gains the issue's shapes and the letters j, y and e (3,168 pairs, 0.48s; 16ee285 disagrees on 30; recorded controls 306 and 117).

Docs. New rules/T4 (the bare letter as an initial) and decisions/T4. P3's Accepted paragraph on a declined marker loses the shape where the own words read differently in the two spellings. M4 and O5 gain examples. P2's sentence about a lowercase v behind a particle had been stale since #721 and is corrected. decisions/S10's Open question is answered. Plus a release-log bullet, the initial tag's docs in _types.py and modules.rst, and AGENTS.md's spelling-sweep sentence.

Reviews. A code review found the clause gap fixed in cf84641. A design-docs review (all eleven axes) found T4's own-words scope unstated, the É wording, the missing grid recipe, the ı twin claim, the unrecorded case-repair move, P2's stale sentence and the maiden-clause numeral move. All are fixed in 52d127e and a1b0b04.

Known gap, not fixed here. It predates this PR: under Policy(lenient_comma_suffixes=False), the strict comma test asks a numeral's shape as written, so john smith, v reads suffix v where JOHN SMITH, V reads given V. It is recorded in decisions/T4 and filed as #727.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.04%. Comparing base (16ee285) to head (d6bebd8).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #726   +/-   ##
=======================================
  Coverage   99.04%   99.04%           
=======================================
  Files          46       46           
  Lines        4585     4586    +1     
=======================================
+ Hits         4541     4542    +1     
  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.