Add the PT9 interlinear converter - #240

Convert the parsed PT9 file set into the extension's analysis layer, the second
slice of the PT9 import pipeline. convertPt9Project is a pure function of the
parsed files, the source project's text, and a timestamp — nothing here calls
papi.

• clusterAnchoring matches each PT9 cluster to a word token by comparing
  spellings in order, because PT9's stored offsets point into a different
  string than our text.
• languageAnalysisBuilder turns one gloss language's clusters into per-token
  and phrase records, marking approved only what PT9 approved.
• analysisMerger combines records for the same token across gloss languages
  into one, keeping genuine disagreements as separate competing records.
• pt9GlossSource looks up a lexeme's gloss text in Lexicon.xml.
• glossLanguageTags turns a PT9 gloss-language value into a language tag we can
  key glosses by.
• lexiconResolver is the seam for linking imported lexemes to the Lexicon
  extension; the stub it ships with links nothing yet.
• bareWordAnalyses converts PT9's wordform lists into analyses that describe a
  spelling rather than any one occurrence, so they carry no links.
• report counts what converted, and what was dropped and why.

index.ts is the converter's public surface: convertPt9Project, its input and
output types, the resolver seam, and the report types. The types each stage
hands the next stay internal to the conversion.

Dropped clusters are counted by reason rather than lost silently, so the report
shows whether a more precise anchoring strategy is worth building. Record ids
derive from the token reference and spelling, so re-importing the same data
produces the same ids. Persistence and the import command follow in the next
slice.

🤖 Generated with Claude Code

This change is