Merge duplicate token analyses on PT9 import (#313)

[alex-rawlings-yyc]

Closes #313.

Summary

• PT9 import merges token analyses with identical content into one, the same way hand-glossing already does. This uses the analysesAreIdentical rule, so surface forms differing only in case also merge. It covers both analyses from interlinear clusters and the unlinked ones from the word-analysis inventories.
• Every link to a dropped duplicate moves to the analysis that is kept, and keeps its own token's surface text, so drift detection stays per token.
• If merging leaves one token with two links to the same analysis, only one is kept: the approved one if either is approved, a rejected one only if both are rejected, otherwise the first.
• The import report has a new merge.identicalPayloadsMerged count. The import modal doesn't show it.

buildBareWordAnalyses keeps its own duplicate check, so skippedExistingIdentical still counts what it did. The merge runs after every other conversion step, so analyses made identical by #324's blank-morpheme stripping will merge too.

Out of scope: a merged analysis shows the surface text of whichever token came first, e.g. a capitalized sentence-initial form; deriving the displayed form from its links belongs to #186. PT9 import still writes one phrase payload per occurrence; that has its own issue.

Test plan

• Unit tests for merging duplicates and moving their links, an unlinked analysis matching a linked one, case variants, the three near-misses that must stay apart (different gloss, breakdown vs. none, different breakdown), and the one-link-per-token rule; 100% coverage
• Import PIA (test-data/pt9-projects/PIA): the four plovs tokens in PHP 1:6-7 end up on three analyses, with the two "catorce / plov s" occurrences sharing one
• The catalog shows that analysis as one row with two uses, and the plovs suggestion dropdown has no duplicate rows

---

This change is 

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 646a5138-8a0a-4825-b391-62cc15db058d

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[jasonleenaylor]

Nit, not blocking: test-data/pt9-projects/README.md:66 still says the plovs pool key gets four competing payloads; with this merge it's three (as your test plan says).

This review was assisted by Claude Opus 5.5.

[alex-rawlings-yyc]

Fixed.

@alex-rawlings-yyc made 2 comments.
Reviewable status: 0 of 10 files reviewed, all discussions resolved (waiting on alex-rawlings-yyc).

[jasonleenaylor]

@jasonleenaylor reviewed 10 files and all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on alex-rawlings-yyc).