When jobboard was enabled in amphorav2, the [task_flow]/max_workers
option was unused, updating a LB with a lot of children could have
failed because of dozens of concurrent connections to the database.
Story 2008860
Task 42381
Change-Id: I5341949513e6eb814655ee8f14a01b061894fc34
(cherry picked from commit c5287e1b08)