What:

Defer placed AFTER Lock() succeeds to avoid unlocking when we don't hold the lock.

Why:

Was running into the following scenario on a multi cluster setup:

1. Cluster1 would run the job and create the lock 👍
2. Cluster2 would attempt to run the job, would see the lock 👍.. however it removed the lock when it quit out 👎
3. Cluster3 would run the job since there was no longer a lock (removed by cluster2) 👎

This fix makes it so if lock-exec fails to create the lock it will not remove any existing locks. Existing locks will only be removed by jobs that create the lock OR if the TTL expires.