Thread leak (and likely memory leak) on Logstash reload

[jakelandis]

When Logstash pipelines are reloaded via the a configuration change, the rufus jobs thread is not shut down.

Profiling via YourKit (filtering live threads by rufus):
After 1 reload:

After another reload:

..and another reload:

The root cause here is that the rufus stop/shutdown is defined as:

    def shutdown(opt=nil)

      @started_at = nil

      #jobs.each { |j| j.unschedule }
        # provokes https://github.com/jmettraux/rufus-scheduler/issue/98
      @jobs.array.each { |j| j.unschedule }

      @work_queue.clear

      if opt == :wait
        join_all_work_threads
      elsif opt == :kill
        kill_all_work_threads
      end

      unlock
    end

...but we are not sending any options to the shutdown method, so the worker threads never get killed. Any objects associated with that thread would considered to be a memory leak. ...pure subjecture is that JDBC drivers may use thread locals to hold cached state which could lead to a decent memory leak over time and volume

[jakelandis]

The fix here is to add the :wait option to the shutdown when reloading a pipeline.

With the fix after a couple reloads: