[6.x] Reset timeout handler after worker loop

[themsaid]

If we have a timeout of 5 seconds, and a job took 4 seconds to run. On the next loop if getNextJob() took 1 second or more to run, the worker will exit before the script reaches registerTimeoutHandler() due to timeout. This will cause the job to not run, yet the attempt will count.

In this PR, we reset the alarm so it doesn't kill the process before the new loop sets a new value for the alarm.

[taylorotwell]

This also introduces the possibility of a totally frozen worker if the code gets stuck on retrieving the next job and the client does not have a proper timeout in place. I'm not sure if any queue drivers are vulnerable to this.

[themsaid]

@taylorotwell I think for at least the drivers recommended for production (redis and sqs), an exception gets thrown if the connection fails. I understand your concern but I think this issue is the cause of the mysterious "Job has been attempted too many times or run too long" error while nothing was attempted.