Ensure non-started jobs are removed from jobs map when stopping

[svanegas]

The problem

Whenever a frame extraction gets cancelled by its request id, there were some cases when a subsequent extract call for the same request id was not reporting any result/error/cancellation.

The cause

The initial cancellation call was fired when the extraction job hadn't started yet, this was causing the job never being interrupted and never getting into a point of triggering a cancellation listener. The cancellation listener is the one in charge of removing the job from the activeJobMap, so since it was never removed, subsequent calls with the same request id were immediately discarded because the job still was in the activeJobMap.

The solution

Add a isStarted property to the FrameExtractJob, which gets set to true right at the beginning of the run method of the Runnable. This property is exposed to ComparableFutureTask, which at the same time can be read from the ActiveExtractJob in order to determine when the job is being cancelled but hasn't started yet.

Visuals

Here's an example of this behavior, modifying a bit the litr-demo we can observe the issue and after the fix.

• Caching was removed for better visualization
• A request id is not random but is uniquely mapped to each frame (based on the timestamp)

Before	After

[izzytwosheds]

Note to self: We should probably do something similar for TransformationJobs as well.