What does this PR do?

The stats-resources task enqueued every active project in the region at the top of every interval, so the worker (maxCoroutines: 1) got an hourly burst and sat idle the rest of the hour.

It now runs on Utopia\Schedule\Scheduler, like ScheduleFunctions, ScheduleExecutions and ScheduleMessages:

• New Appwrite\Schedule\Source\Stats lists a concurrency row plus every project accessed in the last day in this region. Each project gets a Shifted(Interval) trigger with an offset derived from its id, the same spread Source\Functions uses, so its slot is stable across restarts and a region's counts spread evenly across the interval.
• Incremental syncs use the (region, accessedAt) index every minute (or every interval when it is shorter, for test stacks); a full snapshot runs once per interval. Platform DB read cost stays at today's level.
• The payload carries only the project id, since the worker reloads the document. The project listing skips the project subqueries with skipFilters inside the source, as Source\Database does, instead of disableSubqueries(), which mutated the global filter set in the combined scheduler process. A protected subqueries() hook on the task lets Cloud add its own filter names, the same shape as the spreadWindow hook on ScheduleFunctions.
• Dispatch mirrors the sibling tasks: schedule.stats.{reconcile,enqueue,stopped} spans with occurrence.* attributes, and every failure is caught so one bad occurrence never ends the loop.
• Concurrency sampling keeps its cadence (once per interval).

Test Plan

• Unit tests for tests/unit/Schedule/Source pass; the E2E usage suites exercise the task end to end.
• Pint, PHPStan and Rector pass on the changed files.

Related PRs and Issues

Cloud follow-up: appwrite-labs/cloud#5829 overrides the new subqueries() hook and re-points its composer.lock once this merges.

Checklist

• Have you read the Contributing Guidelines on issues?
• Have you added tests?

🤖 Generated with Claude Code