Fix stalled daily account-usage refresh on large hosted Tuist instances

The daily hosted account-usage rollup used to stall on large Tuist instances: it tried to enqueue every per-account refresh job in one multi-row INSERT, which exceeded PostgreSQL’s 65,535 bind-parameter limit once the backlog crossed roughly 7,281 stale accounts. The failed insert meant no per-account jobs were ever enqueued, so the backlog grew every day until the job exhausted its retries and monthly usage data stopped refreshing. Server 1.218.4 now chunks the queue into batches of 1,000 before inserting, so the job completes reliably on large hosted deployments and drains any accumulated backlog on the first run after the upgrade. Behavior is unchanged for smaller instances, which still fit in a single batch.