Fixed flaky-test alerts silently stopping for large quarantine lists

Fixed a scaling issue that caused high-volume flaky-test alerts on the server to stop marking, muting, and recovering tests. The rolling recovery query now aggregates counts inside ClickHouse and processes candidates in batches, so it stays within ClickHouse limits and the alert cursor advances normally. Dormant high-recovery alerts will self-heal on their next tick, and chronically flaky tests will be re-muted correctly.