Cancelled and failed e2e jobs don't clean up resources

[wikkyk]

What steps did you take and what happened:

1. Start an e2e job
2. Cancel it (or let it fail)
3. Notice the created resources linger.

What did you expect to happen: Resources get cleaned up no matter the result.

Anything you would like to add: This makes e2e unreliable as the runner runs out of resources.

Environment: Our internal e2e runners.