Read a title written joined or hyphenated at the end of a name (#1116) - #1125

Closes #1116.

What changes

rules/H5 (the trailing title run) checked a word's shape before asking the vocabulary. The check was _PERIOD_ABBREV: two or more letters, then one period. So a listed title with a period or hyphen inside it was never looked up at the end of a name. It still read as a title at the front and after a comma:

                                 before                               after
Hans Meier Priv.-Doz.            middle 'Meier', family 'Priv.-Doz.'  title 'Priv.-Doz.', family 'Meier'
Meier, Hans Priv.-Doz.           middle 'Priv.-Doz.'                  title 'Priv.-Doz.'
Maxie Bunz Dipl.Psych.           family 'Dipl.Psych.' (2.8.2: suffix) title 'Dipl.Psych.', family 'Bunz'
JORDAN, HERBERT T. LT.COL.       middle 'T. LT.COL.'                  title 'LT.COL.', middle 'T.'
Nguyen Van Anh PGS.TS. (VI)      given 'PGS.TS.'                      title 'PGS.TS.', given 'Anh'

The chain now checks _vocab._TRAILING_TITLE_SHAPE instead: letters, periods and rules/O6's hyphens, starting and ending on a letter and ending in one period. Anything _PERIOD_ABBREV matched still matches, and the front slot's H2 shape is unchanged. The vocabulary lookup after the shape check is the same one as before. I compared each widened spelling with Prof. in the same names: 24 shapes × 3 spellings × 3 orders, fields and reports. All 216 rows read the same. The design-docs reviewer's own grid of 1,248 rows had no differences either.

_HYPHENS moves from _post_rules to _vocab so that O6 and H5 read one set. That means Dipl.‐Ing. with U+2010, the hyphen a word processor writes, chains too. The code review found that gap.

The boundary

The final period is still required, as #1116's first option proposed. Maria Rossi Dott.ssa keeps family 'Dott.ssa'. The Belarusian and Ukrainian д-р, which ends three CLDR renderings, stays the separate decision the issue named.

What the shape reaches is the whole title vocabulary, as the period-marked spelling always did (rules/H5's Accepted bullet). Of the 820 titles in no suffix set, 718 now read as a trailing title when written with a period after them, up from 688. The 30 new ones include whole hyphenated words (Co-Founder., Senior-Judge.), none of which is a name. decisions/vocabulary-exclusions' trailing-position line said a hyphen kept these out; it now records the change.

Measured

Against b1d3d210, using tools/differential/measure.py --base:

• Committed corpora: 2 names move. Both are rules/S3's Accepted rows, which this PR retires. CLDR and GND agreement don't move.
• Local corpora: 13 names move, each a title written joined or hyphenated that leaves a name field: ctgov-officials 3, fec-individuals 3, orcid-researchers 7.
• ORCID agreement: stays 1,478,713 of 1,831,533. The 5 changed records' filers had typed the degree into the family box.
• One move swaps one wrong family name for another: Dr. Fr Rinaldo DIRICCHARDI MUZGA - PhD; Lic.Theol..
• The hyphen commit: re-measured, no change.

Docs

• rules/H5: a Then bullet for the wider shape, with the Dott.ssa boundary.
• rules/S3: the title-chunk Except now covers the end of a name; its Accepted bullet narrows to a word inside the name and to one without a final period.
• decisions/H5: the Should Hans Meier Priv.-Doz. read the listed title at the end, as Hans Meier Prof. does? #1116 entry.
• Head notes: the reversal is noted at the head of Should parser_for(locales.VI) read Nguyen Van Anh as family Nguyen, middle Van, given Anh, and Le Van Anh as family Le? #1108's VI entry and of the Diplom entry.
• Other docs: the root AGENTS.md period gotcha, and the release log.

Reviews

A code review and a design-docs review ran on the first two commits. Their findings are fixed in 5774eebb (hyphens) and e4bd87cb (the reach count, head notes in place of edits to the entry text, and the д-р count). The design-docs reviewer re-checked 5774eebb. A second-round review of e4bd87cb found four errors in its own corrections: the 30 words split 14 + 2 + 14, д-р ends three renderings written without a comma, the head notes needed the "superseded by" form the docs tooling strips, and the reference line didn't say what still keeps the other 102 titles out. All four are fixed in the last commit, which changes only the decisions docs.

🤖 Generated with Claude Code