Fix tests

ml-agents/mlagents/trainers/tests/test_reward_signals.py

         max_steps: 5.0e4
         memory_size: 256
         normalize: false
-        num_update: 1
-        train_interval: 1
+        steps_per_update: 1
         num_layers: 2
         time_horizon: 64
         sequence_length: 64

ml-agents/mlagents/trainers/tests/test_sac.py

         max_steps: 1024
         memory_size: 10
         normalize: true
-        num_update: 1
-        train_interval: 1
+        steps_per_update: 1
         num_layers: 1
         time_horizon: 64
         sequence_length: 16

ml-agents/mlagents/trainers/tests/test_simple_rl.py

         max_steps: 1000
         memory_size: 16
         normalize: false
-        num_update: 1
-        train_interval: 1
+        steps_per_update: 1
         num_layers: 1
         time_horizon: 64
         sequence_length: 32