Set n_workers for dask LocalCluster

 - Explicity set n_workers for LocalCluster creation (through Client).
 If this is not set, for machines with a high CPU count, the number of
 workers will be much larger than 7. In `client.scatter` with
 `broadcast=True`, this can cause hangs due to how dask distributed works.
 It's presumed that this bug is only apparent in more recent runs on
 machines with high CPU counts, although that is not confirmed. But,
 this fix does allow the example to run to completion on a machine
 with CPU_COUNT=128.

examples/run_og_usa.py

@@ -14,8 +14,8 @@
 
 def main():
     # Define parameters to use for multiprocessing
-    client = Client()
     num_workers = min(multiprocessing.cpu_count(), 7)
+    client = Client(n_workers=num_workers)
     print("Number of workers = ", num_workers)
 
     # Directories to save data