Retry transactions when deadlocked

[aaronjensen]

Fixes #63

[albus522]

The latest DJ already does this. When an Exception occurs while locking a job, DJ logs the error, then continues as normal. If the deadlock occurs 10 times in a row, the process will exit with a FatalBackend error.
https://github.com/collectiveidea/delayed_job/blob/master/lib/delayed/worker.rb#L281

[aaronjensen]

@albus522 thanks for that, unfortunately that only covers one half of the problem. The other half is during the delete, that part is still not trapped and runs the risk of delayed job running the job twice when things deadlock. Would you like me to remove the deadlock trapping for the reserve and add it to the delete? Or would it be better to update delayed_job to make delete more robust? This is a pretty specific issue to MySQL from what I know. See all of the conversation in #63, I'm not alone.