step increment moved to _update_policy, fixed exit status issue

config/trainer_config.yaml

     time_horizon: 1000
     lambd: 0.99
     beta: 0.001
-    max_steps: 1.0e5
 
 3DBallHard:
     normalize: true

ml-agents/mlagents/trainers/optimizer/tf_optimizer.py

 
         if hvd is not None:
             adam_optimizer = tf.train.AdamOptimizer(
-                learning_rate=learning_rate * hvd.size(), name=name
+                learning_rate=learning_rate, name=name
             )
             horovod_optimizer = hvd.DistributedOptimizer(adam_optimizer)
         else: