Expose workflow run observability for frontend brain operations dashboards

[chubes4]

Goal

Expose generic workflow/run observability so Intelligence can build brain operations dashboards without scraping Data Machine internals or inferring too much from partial records.

This is runtime substrate work for Automattic/intelligence#814 and the Horse frontend operations surfaces.

Required work

• Provide stable read APIs/CLI output for flow, pipeline, job, processed-item, Action Scheduler, and pending-action state by agent/root/workflow identifiers.
• Include last run, last success, last failure, queue depth, next scheduled work, retry state, and durable error summaries where available.
• Support correlation identifiers that Intelligence brain manifests can map to workflow modes and source lanes.
• Distinguish complete, partial, inferred, and unavailable status.
• Keep output generic; do not encode Intelligence brain semantics in Data Machine.

Acceptance criteria

• Intelligence can consume Data Machine status through stable contracts rather than raw table/action internals.
• Frontend dashboards can show workflow health without requiring wp-admin or CLI access.
• Missing status data is explicit and machine-readable.