Run Scheduler#cache_warm on global thread pool instead of Scheduler's thread pool

[bensheldon]

In single-thread queues, it's possible for the warm_cache operation to have a race condition with newly enqueued jobs that result in the job not being immediately executed because the warm_cache operation has exhausted the executor itself.

Instead, warm_cache should be run on the global thread pool. I don't think this will have side effects.

[bensheldon]

Running on the global thread pool was a bad idea because it's unmanaged. This is a simpler fix:

good_job/lib/good_job/scheduler.rb

Line 227 in bd5cd69

create_task # If cache-warming exhausts the threads, ensure there isn't an executable task remaining