repocleaner: Fix test race caused by too low liveness interval

One of our tests for the repocleaner checks that we skip over any storages that are currently being processed by a concurrently running cleaner already. This test is racy though because we frequently observe that the second cleaner runs even though it should not.

This issue is caused by a liveness interval that's set quite low. This interval dictates how often we update the liveness timestamp in the acquired database lock -- and if a concurrent repocleaner sees that the lock hasn't been updated since the last time period it'll evict the lock and assume that the other repocleaner has died.

Of course, due to the busy nature of our CI runners, we frequently exceed the liveness update period. So let's fix this by bumping it to one hour instead of one second.

Closes #5504 (closed).