Race-condition when starting job schedule

[Pharb]

Problem

Jobs are sometimes no longer executed by configured schedule.

Workaround

After server restart, jobs are running again as expected.

Cause

The suspected problem is in SchedulePing.checkActiveSchedule(). When startAllJobs() takes longer than interval, a new iteration of checkActiveSchedule() will run and concurrently run startAllJobs(), since startedJobs is not yet set to true. This causes stop() to be run on the JobScheduler which sets the jobExecutor to stopped. This will prevent the jobExecutor to decrease its running counter at the end of its execute function.

To solve, either the stopped flag in the jobExecutor needs to be cleared during the start of the JobScheduler, or it needs to be made sure that the start function is only run once. Either way it would make sense to only allow one execution of checkActiveSchedule to run in parallel.

Reproduce

See #603 for a test case demonstrating this race-condition.