Improve linking tasks retry handling
Since !117 (merged) we only allow a single retry for all tasks, which might not be enough for linking tasks.
For now, we can only have a single requeue, which will fail if the media we're trying to link hasn't been analyzed yet.
Ideally, we should have a per-task number of retries, so instead of storing
retry_count and ensuring they are less than or equal to the max number of retries, we could just store the number of retries left for this tasks, and configure this when creating the tasks.