Jobs can unlock mutexes they don't own

[pik]

This happens with perform_at or when a job runs longer than the key_expire time, this is also explicitly advised against in the redis docs.

[mhenrixon]

Am I understanding it correctly that the new job gets unlocked by a newer job is there something I am missing?

[pik]

So at the moment jobs do not check if they were the ones which originally set a mutex when unlocking -- meaning there are conditions where the job will delete a mutex that another job has set (if it's own has already expired).

[mhenrixon]

Closed by #96