Add ActiveScheduler.config.check_jobs_descend_from_active_job

[AMHOL]

When set to false, this will skip the check in ActiveScheduler::ResqueWrapper.wrap which checks whether the job class descends from ActiveJob::Base, this means that we no longer need to call class_name.constantize, which should mean that we can skip loading the whole Rails environment for the resque scheduler rake task and save some memory.

[AMHOL]

From #20

I had also considered adding a configurable guard proc instead of this, to allow for custom matching to wrap jobs rather than it being all-or-nothing, but I couldn't come up with a good name and it got messy when trying to resolve the queue name, which I ended up adding an additional queue_name_resolver config for. I could spend some more time on that approach if you prefer it to this. The config would have ended up being something like:

ActiveScheduler.configure do |config|
  config.guard_with = ->(class_name) { class_name.constantize <= ActiveJob::Base }
  config.queue_name_resolver = ->(class_name) { class_name.constantize.queue_name }
end

With this approach, we would lose the ability to raise an ActiveScheduler::QueueConfigMissingError exception when the queue name is not configured and we don't check the job class descends from ActiveJob::Base

[coveralls]

Coverage remained the same at 100.0% when pulling 0db79fc on AMHOL:feature/check_jobs_descend_from_active_job-config into ee3a38c on JustinAiken:master.