Adapt large workers to manager large origins

"Normal" repositories have been ingested so we can let the worker manage the "large"
ones as well with not too much concurrency.

Related to T4400