TenantOplogBatcher should transition to complete when the next batch callback is called with error status

_batchRequested is not reset to false if the scheduled callback is called with a non-OK status. I think the convention of the TenantOplogBatcher is that we should reset _batchRequested to false and check for shutdown when it becomes false. This is a problem similar to SERVER-54509 but when the callback is called with an error status. This would result in server hang on stepDown when a tenant migration recipient primary is trying to join the TenantOplogApplier.