In current implementation --abort-on-sla-failure affects
only currently running workload and allows to run next
subtasks/workloads after sla failure.
* Change status of task to SOFT_ABORTING after sla failure.
This allows to consume results from pending iterations
and abort task before running next subtask/workload.
Change-Id: I1498b62fd6929101ff4d45ca78e185f124ab7b89