meng and lac join the ambiguous suffix acronyms (#540)

[derek73]

meng (MEng) and lac (LAc) join SUFFIX_ACRONYMS_AMBIGUOUS. Both are borne as names in the one slot the suffix vocabulary claims, the last word of a name, and the unambiguous claim cost the family name: wang meng has read suffix meng with no family since 2.0.0, and since #521 rendered Wang MENG. Two lines of data; rules.md#S2's words-to-spare and written-case reading does the rest, exactly as it does for ma and ba.

Closes #540

What changes

• A bare trailing Meng or Lac with no words to spare is the family name. wang meng → given wang, family meng, with a suffix-or-name report (every 2.x release: suffix meng, no family; 2.3.0 reported given-or-family, the wrong fork; 1.4.0 read the family, unflagged). tran lac moves the same way; Wang, Meng and Smith, meng get 1.4.0's given name back after a family comma; under FAMILY_FIRST, Wang Meng reads given Meng. The render follows: Wang Meng, Tran Lac.
• With words to spare the credential stays, now reported: john smith meng → suffix meng (John Smith MENG), nguyen van lac → suffix lac. Capitals keep it in a mixed-case name (John Smith MENG, Smith, MENG).
• The cost is the one ma already carries, and it falls on the conventional spellings: in a name written in more than one case S2 reads a member written neither in capitals nor all lower as the name even with words to spare. john smith MEng → middle smith, family MEng; Smith, John MEng → middle MEng; Smith, MEng → given MEng; John Smith (MEng) → nickname. Every release read suffix MEng for all four. A credential run ending in one of them goes the same way (Mary Jones PhD MEng → middle Jones PhD, family MEng; John Smith MEng PhD → family MEng, suffix PhD), S2's declined pick stopping the walk as it does for Doe, John MA Ma. Behind a suffix comma the run is no longer wholly suffix-shaped, so the comma re-reads as a family comma in every case form (John Smith, PhD MEng → given PhD, middle MEng, family John Smith; a title-listed front word reads as a title with no report, Jane Doe, MS LAc), the path john smith, phd ma already takes; a second comma before the run keeps the credential (John Smith, PhD, MEng). Wang M.Eng. → family M.Eng.: the period gate counts a member as unambiguous only written one period per letter, and these are the first members whose dotted spelling is chunked (accepted and recorded, decided; John Smith M.Eng. keeps the suffix). Both the comma re-read and the period-gate question are Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544. The Title-case Nguyen Van Lac → family Van Lac is the gain for a Vietnamese name.
• Declined with measurement (decisions.md): removal (loses the bare credential outright, john smith meng → family meng) and masks meng → MEng / lac → LAc (the mask clause is role-free, so it would re-spell the restored surnames Wang MEng, MEng Li). edd and ded stay unmarked: borne as given names, not the trailing slot; a trailing bearer reopens them.

Measured

No corpus name carried either word before this change, so every mover is one of the fourteen shape-tagged case rows this PR adds, nine with no comma and five with one (full name before it, lone word before it, a credential run behind it). Classified under eight fix(#540) ledger rules at each 2.x baseline and four at 1.4.0; the gate exits 0 at all five with nothing unexplained and the radar tier unchanged.

baseline	intentional diffs	#540 names
1.4.0	545 → 552	john smith MEng, John Smith MEng PhD, Smith, John MEng, John Smith, PhD MEng, john smith, phd meng, Nguyen Van Lac (+ nguyen van lac on initials, fix(#385/#402))
2.0.0 / 2.1.0 / 2.2.0 / 2.3.0	509 → 522 / 421 → 434 / 275 → 288 / 163 → 176	the thirteen rows that move or gain a report

Role fields moved on zero pre-existing corpus names. Recount by decisions.md#vocabulary-collisions' recipe: suffix_acronyms 608 with 7 ambiguous (5 before).

Docs, same PR

rules.md#S2 gains four example lines; decisions.md#suffix-acronym-collisions gets the dated decision (balance criterion, every comma path, the costs), the two Declined alternatives with their recipes, and the measurement; the R4 Excluded block's "to be filed" line goes (answered), with a dated supersession note on the sentences #540 overtook and the lac bearer corrected in place (a Vietnamese given name, the trailing word in native order); one release-log bullet with every before/after.

Review

Sonnet implementer, Opus reviewers: spec compliance (an independent replay of every edit byte-identical to the tree), code quality with seven planted mutants all killed (the one survivor deletes a check), two re-reviews of fix rounds, the design-docs reviewer on docs/design/, a spec-and-claims reviewer on the wheels, a whole-branch review whose silent-failure sweep ran 550 texts under three orders against 2.3.0 and matched every bare-spelling mover to the pinned ma fork of the same shape, and a /pr-review-toolkit round (code, tests with 20 planted mutants, comments against all five wheels). Review found and fixed: the comma paths unpinned (three rows and two ledger rules added), a removal recipe that raised on the tree, a liveness argument where the entry's criterion asks for balance, Lac misdescribed as a surname, a Title-case Vietnamese name filed as a cost, stale standing counts in the fix(#385/#402) ledger prose, four unstated movers (the credential run, the comma re-read behind a suffix comma, the Title-case Nguyen Van Lac, the chunked dotted Wang M.Eng.), a release-note claim false on four wheels (the pre-2.3 comma joiner), a ledger rule whose anchors could be dropped unnoticed, and a suffixes.py comment false since #289.

Verification

• uv run pytest -q: 9987 passed, 324 skipped, 3 xfailed
• uv run mypy: clean (111 files); uv run ruff check .: clean
• sphinx-build -b doctest and -b html: 0 failures, no warnings; python -m doctest README.rst: clean
• tools/differential/compare.py at 1.4.0 / 2.0.0 / 2.1.0 / 2.2.0 / 2.3.0: exit 0; corpus_shapes.jsonl 223 → 237, corpus_rules.jsonl 414 → 420

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.78%. Comparing base (120033b) to head (f8de0ac).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #543   +/-   ##
=======================================
  Coverage   98.78%   98.78%           
=======================================
  Files          45       45           
  Lines        3703     3703           
=======================================
  Hits         3658     3658           
  Misses         45       45           

☔ 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.