Recognize maiden-name markers like "née" and "geb." (Jane Smith née Jones)

[derek73]

The maiden attribute is currently populated only by delimiter routing (e.g. moving parenthesis into maiden_delimiters for the "Jane Smith (Jones)" convention). But the dominant convention across languages — including English-language biographical text — is a marker word before the former surname, which the parser currently mangles into the middle name (v1.3.0):

>>> hn = HumanName('Jane Smith née Jones')
>>> hn.middle, hn.last, hn.maiden
('Smith née', 'Jones', '')      # want: middle='', last='Smith', maiden='Jones'
>>> HumanName('Anna Müller geb. Schmidt').middle
'Müller geb.'

Proposal: a maiden_markers constants set. When a marker is found, it consumes itself plus the following name piece(s) into maiden_list — structurally similar to how prefixes join forward. Candidates:

• née, né, nee — French; standard in English text
• geb., geborene — German; geboren Dutch
• roz., rozená — Czech/Slovak
• z domu — Polish (multi-word marker; the mechanism must handle two-token markers)
• урожд., урождённая — Russian (would land in the non-Latin constants effort, Provide constants in non-Latin scripts (Cyrillic, Greek, Arabic, Hebrew) #269)

Needs vetting before inclusion: Scandinavian f. (født/född) — indistinguishable from the initial "F.", so it likely can't be a default. Out of scope: Hungarian -né (morphological suffix on the husband's name, Kovácsné = "wife of Kovács" — would need ending-detection like the patronymic regexes, not a marker word) and Spanish/Portuguese conventions (surnames aren't replaced at marriage; no marker exists).

Design questions: how many pieces after the marker belong to maiden (probably all remaining non-suffix pieces, mirroring prefix-join behavior); and whether nee without the accent is safe as a default (it's also a rare surname). Related: #269.

[derek73]

The vocabulary side of this is now settled on the 2.0 core-foundation branch; what remains is the consuming rule in the parse pipeline.

nameparser/config/maiden_markers.py ships a MAIDEN_MARKERS set, single-sourced into the 2.0 default lexicon the same way titles/suffixes/conjunctions are. The 1.x parser does not read it. Contents (stored normalized — lowercase, no periods, so geb. matches):

• née, né, nee — French; standard in English text
• geb, geborene — German; geboren — Dutch
• roz, rozená — Czech/Slovak
• урожд, урождённая / урожденная, урождённый / урожденный — Russian

A few decisions here update this issue's text:

• Russian is in the default set, not staged to Provide constants in non-Latin scripts (Cyrillic, Greek, Arabic, Hebrew) #269 / the RU locale pack. The locales design sorts vocabulary by collision risk, not script: verified non-colliding data belongs in the default lexicon. These are participles/abbreviations that can't appear as a name token in any Cyrillic-script language, so they're safe everywhere. (Scandinavian f. is excluded by the same rule — it collides with the initial "F.")
• Both ё and е spellings are listed because str.casefold() doesn't fold ё→е and running Russian text routinely writes е.
• Masculine forms included — rare (a man taking his spouse's surname) but grammatically valid and equally collision-free.
• nee without the accent stays a default. The rare-surname collision only bites when the token is followed by another name piece; if the differential harness shows misfires, removing a default entry before 2.0.0 costs nothing.
• z domu is deliberately absent pending the pipeline decision on multi-token markers. The set could hold a spaced entry today, but that would be vocabulary with no consuming rule.
• Hungarian -né and Spanish/Portuguese conventions stay out of scope as proposed above.

Still open, for the pipeline stage of this issue: how many pieces after the marker land in maiden (leaning: all remaining non-suffix pieces, mirroring particle joining), and the multi-token matching mechanism that would unlock z domu.

[derek73]

Follow-up: the full Scandinavian participles født (da/nb), fødd (nn), and född (sv) are now in the default set — the collision that vetoed Scandinavian was specific to the abbreviation f., which remains excluded. No ASCII variants (fodt/fodd): unlike nee in English, dropping the diacritic isn't standard practice in Scandinavian text.

[derek73]

Shipped in 2.0.0: née/nee, geb./geborene/geboren, roz./rozená, the Scandinavian participles, and the Russian forms all route the following name to the new maiden field, including multi-piece consumption ("née van der Berg"). One residual: the Polish two-token marker "z domu" is descoped pending multi-token vocabulary matching — tracked on #291 with the rest of that mechanism.