In Job implementations handle failures of steps

[woksin]

For instance when starting a job that has multiple steps, if some of these steps even fail to start then when the job completes it should be able to handle the failed steps in a way (starting another job, creating a recommendation, etc...)