Error handling for jobs with job chains on Agents should work

Current Situation

• When using a PowerShell job in a job chain then error handling does not work if the job is configured to fail in case of output being written to stderr by <job stderr_log_level="error"/>. The same behavior applies to shell jobs that are executed with Agents.
• Consider the attached sample job chain from the JS-1734.zip archive
  • the job in the first node will raise an error
    • the job is configured for stderr_log_level="error"
• When running the attached sample job chain then
  • an order will not leave the job chain with the "error" node but with the "success" node.
  • the order log shows the following output:

    2017-10-12 09:53:35.489+0200 [info]   
    2017-10-12 09:53:35.489+0200 [info]   Order powershell_order - Protocol starts in C:/ProgramData/sos-berlin.com/jobscheduler/apmacwin_4444/logs/order.JS-1734,powershell_chain.powershell_order.log
    2017-10-12 09:53:13.792+0200 [info]   SCHEDULER-893  Standing_order is 'active' now
    2017-10-12 09:53:35.505+0200 [info]   (Task JS-1734/powershell_job1:431) SCHEDULER-842  Task is going to process Order JS-1734/powershell_chain:powershell_order, state=run1, on JobScheduler 'http://apmacwin:40444'
    2017-10-12 09:53:35.505+0200 [info]   (Task JS-1734/powershell_job1:431) 
    2017-10-12 09:53:35.505+0200 [info]   (Task JS-1734/powershell_job1:431) Task JS-1734/powershell_job1:431 - Protocol starts in C:/ProgramData/sos-berlin.com/jobscheduler/apmacwin_4444/logs/task.JS-1734,powershell_job1.log
    2017-10-12 09:53:35.651+0200 [info]   (Task JS-1734/powershell_job1:431) SCHEDULER-726  Task runs on remote scheduler http://localhost:4445
    2017-10-12 09:53:35.651+0200 [info]   (Task JS-1734/powershell_job1:431) SCHEDULER-918  state=starting (at=never)
    2017-10-12 09:53:41.264+0200 [info]   (Task JS-1734/powershell_job1:431) hello world
    2017-10-12 09:53:41.583+0200 [ERROR]  (Task JS-1734/powershell_job1:431) 
    2017-10-12 09:53:41.583+0200 [ERROR]  (Task JS-1734/powershell_job1:431) echo "hello world"
    2017-10-12 09:53:41.583+0200 [ERROR]  (Task JS-1734/powershell_job1:431) 
    2017-10-12 09:53:41.583+0200 [ERROR]  (Task JS-1734/powershell_job1:431) write-error "an error occurred"
    2017-10-12 09:53:41.583+0200 [ERROR]  (Task JS-1734/powershell_job1:431)          : an error occurred
    2017-10-12 09:53:41.583+0200 [ERROR]  (Task JS-1734/powershell_job1:431)     + CategoryInfo          : NotSpecified: (:) [Write-Error], WriteErrorException
    2017-10-12 09:53:41.583+0200 [ERROR]  (Task JS-1734/powershell_job1:431)     + FullyQualifiedErrorId : Microsoft.PowerShell.Commands.WriteErrorException
    2017-10-12 09:53:41.583+0200 [ERROR]  (Task JS-1734/powershell_job1:431)  
    2017-10-12 09:53:41.590+0200 [info]   (Task JS-1734/powershell_job1:431) SCHEDULER-843  Task has ended processing of Order JS-1734/powershell_chain:powershell_order, state=run1, on JobScheduler 'http://apmacwin:40444'
    2017-10-12 09:53:41.590+0200 [info]   set_state success
    2017-10-12 09:53:41.598+0200 [info]   SCHEDULER-944  End state reached - order will be repeated at 2038-01-19 03:14:07.000Z with state=run1
    2017-10-12 09:53:41.598+0200 [info]   SCHEDULER-962  Protocol ends in C:/ProgramData/sos-berlin.com/jobscheduler/apmacwin_4444/logs/order.JS-1734,powershell_chain.powershell_order.log
    

Desired Behavior

• Error handling should work for jobs with Agents that use <job stderr_log_level="error"/> in the same way as for jobs with a Master.