JITSU-243: Show Meta Reverse ETL audience and conversion results

[absorbb]

JITSU-243

Meta Reverse ETL currently shows API acceptance counts without showing what Meta reports about the target. Add a separate current-results panel to run logs, with a read-only manual refresh: approximate audience size and processing/ad eligibility, an eligible estimated match range, and per-web-event Event Match Quality and Additional Conversions Reported from the Dataset Quality API.

Results describe the whole audience/dataset, including other senders; they are not matching or attribution results of the selected historical run. Missing, delayed, privacy-limited and invalid metrics stay distinct from zero. Reporting failures explain token/dataset permissions or Events Manager opt-in and cannot change delivery or saved state.

A match estimate requires the latest successful completed exclusive managed-mirror snapshot and an unambiguous source-row denominator. The runner adds optional sealed snapshot counts to aggregate task statistics. Existing audiences, failed/pending runs, older runs without counts, empty/excluded/duplicate/fan-out snapshots and unreliable size estimates receive an explicit unavailable reason. No snapshot payloads are downloaded, and these aggregates do not affect delivery/recovery authorization.

Independent of #1559. No automatic replay, audience provisioning, state resets or schema changes.

Validation: 21 provider reporting tests, 309 console unit tests plus the retargeting-cache regression, 15 real Postgres/ClickHouse reporting integration tests, 486 runner tests (1 skipped), all three package typechecks, runner build, changed-console lint and formatting. Live Meta result availability still depends on reporting permissions, data processing and privacy thresholds.

Deploy updated retl-runner first, then console. Subsequent runner executions supply snapshot counts; older attempts remain readable without a match estimate. Console deployment remains manual unless explicitly requested otherwise.

[jitsu-code-review]

Reviewed the Meta reporting endpoint, server-side target/snapshot validation, runner metrics, and console presentation.

Finding:

• Snapshot eligibility can publish a match rate for a source set that contains excluded or multi-member rows.

[jitsu-code-review]

Reviewed the Meta destination-results flow across runner metric persistence, server-side target validation, Graph API result handling, and console rendering.

Finding:

• The displayed snapshot timestamp can be refreshed by later delivery observations, rather than reflecting when the snapshot was sealed.

[jitsu-code-review]

Reviewed the Meta destination reporting flow, including runner snapshot eligibility, workspace/target scoping, credential-safe Graph reads, API response validation, and console caching/rendering.

No actionable bugs, security issues, or user-visible regressions found.