Should bon, dí and heer stay particles? None is attested as one #766

bon and dí have been in the particle list since the earliest history (bon in the 2011 svn import, dí in fc2ea89, 2012). Nothing records which names they come from, and #749 found no particle use of either. Both are borne as names (counts recorded on their entries in nameparser/vocab/und.toml, 2026-10-19):

• Bon: 1,123 US Census 2010 surname bearers, 274 male SSA births, and 102 Wikidata family-name bearers (mostly French and Italian).
• Dí: 7 Wikidata given-name bearers, 5 of them PRC citizens.

heer (Dutch "lord", now in nameparser/vocab/nl.toml as a rank) came in with #18's German/Dutch additions (3e14ea2), and its only source is that issue's test name Johan de heer Van Kampen (family de heer Van Kampen). That name looks like two real ones run together: Johannes de Heer appears in a lexicon of Dutch Protestant history ("Heer, Johannes de"), and Van Kampen is an ordinary surname (Nico van Kampen). The full string occurs elsewhere only in another name parser's test data (full_name_splitter 1.0.3). No real name supports the particle reading.

As particles these words join the next word into the family name (Marie Bon Durand → family Bon Durand).

Proposal: remove all three from the particle vocabulary, recording each as excluded = { particle = … } with this evidence. Then read the snapshot moves.

Part of #747.