Check interval_reference option when a job times out or fails

[s-aebischer]

Describe the change

When a job times out or fails a retry_timeout can be specified. The retry_timeout is added to the current timestamp.
The scheduler should check the interval_reference option and schedule the new job by adding the retry_interval to either the start time of the job or the current timestamp.

Current situation

When a job failed or ran into a timeout the new execution time is always calculated by adding the retry_interval to the current timestamp.

Should

The scheduler should check the interval_reference and add the retry_interval to either the start time of the job or the current timestamp