fix(#562): a particle chain unsettles a comma credential run

[derek73]

Closes #562.

What changes

C1's run test had a shortcut. When every ambiguous-class member of the part after the comma is listed and written in capitals in a mixed-case name, it skipped the flip and left the part to the family-comma path, which promised to read it as a credential run. That promise fails when two particles stand next to each other. Group chains them into one particle run, assign reads the part as name text, and P6 attaches the chain to the family.

Input	2.3.0	master	this PR
John Smith, PhD DO DO	given PhD, family DO DO John Smith	same	given John, family Smith, suffix PhD DO DO, reported
John Smith, DO DO DO	given DO, middle DO DO	same	suffix DO DO DO
John Smith, PhD vd DO	given PhD, family vd DO John Smith	same	suffix PhD vd DO
John Smith, PhD DO	given PhD, family DO John Smith	suffix PhD DO, silent	unchanged from master

1.4.0 reads every one of these the way this PR does.

This is #562's fix 1. The settled check gives way when two particles stand side by side, and the count of name words before the comma reads the run instead. rules.md#S2 already says the capitals don't decide a chained member, so S2 stays as written. Fix 2 would have needed S2 amended.

The check is any two adjacent particles, not only a class member behind one. vd is a particle and an unambiguous suffix word, so PhD vd DO and MA vd vd misread the same way. Segment runs before classify, so the check uses classify's own particle test (_normalize(text) in lexicon.particles). It is asked only while the run is still settled, so other comma names pay no extra frame.

Measured (against master 0eadede, details in the decisions.md entry)

• Differential corpus: 0 of the 1,453 shared names move. The gate exits 0 at all five baselines. Its only movers are the two names this PR adds, under a new fix(#562) rule in the four 2.x ledgers.
• Settled grid (test_properties.py, 5,580 texts): 15 move, 10 of them role moves. The pinned _SETTLED_EXCEPTIONS are gone, and the count and both recorded negative controls are re-measured.
• Wider grid (4,389 texts over PhD/MA/Ma/DO/Do/do/vd/van/Jr/MD/Ms): 30 move, 17 of them role moves. Every mover now reads given John, family Smith, with the part as suffix.

Decided in review (Derek, 2026-10-01)

• The new reports on runs that already read correctly are kept. Examples are John Smith, vd DO, MD DO DO and Esq. DO DO. These aren't names anyone writes on purpose, so a report is the right signal. Recorded in the decisions.md entry and pinned by a case row.
• No correct reading exists for John Smith, PhD DO. DO or PhD DO y DO, so this PR leaves them as they are.
• John Smith, MA MA reporting is wanted, because nobody repeats MA at the end of their name on purpose. rules.md#C1's sentence that such a run "reads whole in silence" is therefore the inaccurate part, not the behavior. That sentence predates Why does John Smith, PhD DO DO read given 'PhD' and family 'DO DO John Smith'? #562, so it gets its own docs PR rather than this one.
• Left open, as Why does John Smith, PhD DO DO read given 'PhD' and family 'DO DO John Smith'? #562 asked:
  • Smith, PhD DO DO has one name word before the comma, so it stays in listing form: given PhD, family DO DO Smith.
  • John Smith, PhD van der uses non-member particles, which never reach the run test.

Review

• Rounds: a design-docs review ran all eleven axes and a code review ran on the first commit. The fix commit was reviewed again, and its one finding is fixed in a follow-up commit.
• Corrections from review:
  • P2, not S2, does the join, and P2 is now in C1's interacts:.
  • Two wrong claims in the release bullet are fixed.
  • The report-only movers in decisions.md are now listed by example rather than as a complete set.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.87%. Comparing base (0eadede) to head (82f2388).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #569   +/-   ##
=======================================
  Coverage   98.87%   98.87%           
=======================================
  Files          45       45           
  Lines        3985     3990    +5     
=======================================
+ Hits         3940     3945    +5     
  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.