Expose rule false positive ratio to rule hit table.

[apognu]

The false positive ratio must be computed from the latest outcome of closed cases where a rule appears, this required adding the org ID to case events, ingesting some of the case events, and performing a separate query.

A migration adds the org ID, an appropriate index and backfills the data for the case_events table.

Not tested on production data yet.

---

Note

Adds false-positive ratio to the rule hit table by exporting latest closed-case outcomes per rule to rule_hit_outcomes, requiring org_id on case_events and updating event creation paths.

• Analytics queries:
  • Extend analytics.RuleHitTable with rule_id and false_positive_ratio; add FalsePositiveRatio type.
  • RuleHitTable endpoint now joins aggregated false positives from rule_hit_outcomes (latest outcome per case).
• Data export pipeline:
  • New AnalyticsCopyCaseEvents to copy latest outcome_updated events for closed cases into rule_hit_outcomes (partitioned by org/year/month/trigger_object_type).
  • Export job now runs exportCaseEvents in parallel with other exports and watermarks via AnalyticsGetLatestRow.
  • Add WithCtesRaw to build CTEs with raw placeholders.
• Models/DB:
  • Add OrgId uuid.UUID to models.CaseEvent, CreateCaseEventAttributes, and DB adapter; batch insert now writes org_id.
  • Migrations: add org_id column to case_events, backfill from cases, and create partial index idx_case_events_by_org on (org_id, event_type, created_at).
• Usecases/Repositories:
  • Update all case-event creations to populate OrgId (case create/update, assign/unassign, snooze/unsnooze, comments/tags/files, decision review, SAR, screenings, rule snooze, escalate, mass update).
  • Change UpdateDecisionsWithEvents to accept orgId and propagate across callers.

^{Written by Cursor Bugbot for commit 30d0fe4. This will update automatically on new commits. Configure here.}

[Pascal-Delange]

I created the index manually (⚠️ and created the org_id column and backfilled it for only outcome_updated events, so need to roll this back before deploying prod !!! ⚠️ ), so we can now test the query in prod.

[Pascal-Delange]

Makes sense like this 👍
As discussed in DM, check the limit placement, I think as it is we can lose some decisions/rules due to waterparking/pagination/limit too late.

[Pascal-Delange]

LGTM!