Match Cluster's use of regular methods for scaling

Instead of having coroutines used for `scale_up` and `scale_down`, use
regular methods in their place. Move the coroutines into internal spec.
This breaks the API of dask-drmaa. However is technically more correct
given the Cluster API's current expectations.

dask_drmaa/core.py

@@ -290,12 +290,15 @@ def stop_workers(self, worker_ids, sync=False):
             get_session().synchronize(worker_ids, dispose=True)
 
     @gen.coroutine
-    def scale_up(self, n, **kwargs):
+    def _scale_up(self, n, **kwargs):
         yield [self.start_workers(**kwargs)
                for _ in range(n - len(self.workers))]
 
+    def scale_up(self, n, **kwargs):
+        self.scheduler.loop.add_callback(self._scale_up, n, **kwargs)
+
     @gen.coroutine
-    def scale_down(self, worker_ids):
+    def _scale_down(self, worker_ids):
         worker_ids = list(set(worker_ids))
 
         for wid in worker_ids:
@@ -312,6 +315,9 @@ def scale_down(self, worker_ids):
 
         logger.info("Scaling down workers %s", worker_ids)
 
+    def scale_down(self, worker_ids):
+        self.scheduler.loop.add_callback(self._scale_down, worker_ids)
+
     def close(self):
         logger.info("Closing DRMAA cluster")
         self.stop_workers(self.workers, sync=True)