fix: defer bulk-import re-ID until all detections terminal

[JasonWildMe]

Summary

• Bug: Bulk imports fire per-encounter identification immediately after each detection callback, before sibling encounters finish detection. Intra-bulk matches are missed on first pass because WBIA's matching index doesn't include annotations from unfinished siblings.
• Fix: Branch processCallback for bulk imports: defer identification until all ImportTask MediaAssets have terminal detection status (complete|pending|error), then fire one bulk-final ID pass over ALL eligible annotations, claimed atomically via JVM-global registry + persistent Task flag for restart survival.
• Non-bulk path unchanged — refactored into helpers but identical behavior.

Architecture

Detection callback arrives at processCallback()
   │
   ├── Non-bulk? → existing per-callback ID kickoff (unchanged)
   │
   └── Bulk import (importTaskId in task params)?
         │
         ├── Not all detections terminal? → return (defer)
         │
         ├── Already fired (persistent flag)? → return
         │
         ├── Another thread won claim? → return
         │
         └── Claim won → mark persistent flag
              → enumerate ALL ImportTask annotations
              → filter: matchAgainst && validForIdentification
              → group by encounter, kick off ID per encounter

Files changed

File	Change
BulkFinalIdentificationRegistry.java	NEW — JVM-global ConcurrentHashMap single-winner claim
Task.java	ADD persistent bulkFinalIdentificationFired Boolean
package.jdo	ADD JDO mapping for new Task field
ImportTask.java	ADD isAllDetectionTerminal() helper
IBEISIA.java	REWRITE identification block + 5 extracted helpers

Test plan

• 24 new unit tests (registry atomicity incl. 50-thread race, bulk task detection, encounter grouping, terminal status)
• Full mvn test passes (11 pre-existing errors in Settings/ShepherdState tests unrelated to this change)
• mvn compile clean
• Independent code review by Codex: "Code is sound, ready for PR"
• Manual verification on test install with ≥3-encounter bulk import — log line to watch: INFO: fireBulkFinalIdentificationIfReady: ImportTask X -> ID on N annotations across M encounters

🤖 Generated with Claude Code

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 51.61%. Comparing base (40ea242) to head (5d9f55b).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1552   +/-   ##
=======================================
  Coverage   51.61%   51.61%           
=======================================
  Files         308      308           
  Lines       11976    11976           
  Branches     3856     3847    -9     
=======================================
  Hits         6182     6182           
- Misses       5507     5514    +7     
+ Partials      287      280    -7     

Flag	Coverage Δ
backend	51.61% <ø> (ø)
frontend	51.61% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[goddesswarship]

@naknomum For your approval when you return!