This will cancel all jobs with one query. Instead of cancelling one by one.
In case that some action gets scheduled multiple times e.g. due a race condition
it will remove all jobs.
We use action_scheduler_maximum_execution_time_likely_to_be_exceeded filter for reading remaining execution limit
tracked in the Action Scheduler runner.
Then we use the remaining limit minus safety margin for the daemon execution.
[MAILPOET-4274]
In afterProcess callback we trigger a new runner when there are still some scheduled tasks
to process. Before this change we did this for every action scheduler run and it was causing that
we triggered more HTTP requests than needed.
[MAILPOET-4274]
