Document FAMILY_FIRST_GIVEN_LAST, and fix the name-order example it exposed

[derek73]

FAMILY_FIRST_GIVEN_LAST appeared in the docs only as reference-table entries — the API entry in modules.rst, one clause in the Policy field table, and a changelog bullet. Nothing showed what it does to a name. Since it is keyed to no script and has no locale pack, an explicit Policy is the only way to reach it, so the guide was the one place a reader could have found it.

Writing that up surfaced a problem in the example already there.

The old example was inert

customize.rst demonstrated FAMILY_FIRST with "Nguyen Van Minh" — a Vietnamese name, while modules.rst names Vietnamese as the exemplar for FAMILY_FIRST_GIVEN_LAST. The two pages disagreed on the same name.

It also could not distinguish the two constants:

Nguyen Van Minh
  FAMILY_FIRST     family='Nguyen'  middle=''  given='Van Minh'
  FF_GIVEN_LAST    family='Nguyen'  middle=''  given='Van Minh'

Van collides with the Dutch particle van in particles_ambiguous, so the vocabulary layer joins it forward before the positional layer ever sees three tokens. The doctest passed, but it was exercising particle-joining rather than name_order, and reported the Vietnamese given name as "Van Minh" rather than "Minh".

Remove the collision and the constants separate:

Tran Quoc Toan
  FAMILY_FIRST     family='Tran'  middle='Toan'  given='Quoc'
  FF_GIVEN_LAST    family='Tran'  middle='Quoc'  given='Toan'   ← correct Vietnamese

Changes

1. FAMILY_FIRST moves to a Hungarian name — what modules.rst already claims for the constant, and free of the collision. The middle field is now shown so the assignment order is visible.
2. FAMILY_FIRST_GIVEN_LAST gets the Vietnamese name, contrasted against FAMILY_FIRST on that single name, with a sentence of naming background first and a note that the order is keyed to no script (Latin carries none) and has no vn pack yet (Wrong parsing of vietnamese names #146).
3. The collision is recorded as prose plus a unit test rather than a second doctest block, per the lean-docs rule.

It extends the existing name-order section rather than adding a sibling — the constant only means anything in contrast with FAMILY_FIRST, and a separate section would duplicate the comma-precedence paragraph that follows.

Verification

• uv run pytest — 3066 passed, 20 skipped, 11 xfailed
• uv run sphinx-build -b doctest docs — 223 tests, 0 failures (up from 218; +5 matches the new block's 5 statements)
• uv run sphinx-build -b html -W --keep-going docs — build succeeded, so the internal section link resolves
• The doctest harness was checked against a deliberately wrong expected value first, and reported the failure — the passing run is not inert

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.48%. Comparing base (5a16424) to head (7050f1e).
⚠️ Report is 3 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #347   +/-   ##
=======================================
  Coverage   98.48%   98.48%           
=======================================
  Files          41       41           
  Lines        2845     2845           
=======================================
  Hits         2802     2802           
  Misses         43       43           

☔ 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.