Operator health view of all collectors in a single response

One round-trip for the collectors dashboard.

Bundles the three queries an operator runs manually when checking 'are collectors healthy': • per_source: 24h aggregate (ok / err counts, last_run timestamp, last_status, last_error, total_new) per collector source • checkpoints: latest successful checkpoint per source — shows whether incremental watermarks are advancing • db_counts: total rows in the destination tables, so a healthy run that adds 0 records (incremental, nothing changed) doesn't get misread as broken