Correct actor learning rate (#181)

According to the original DDPG paper, it's 1e-4.

Author: katanachan

rllab/algos/ddpg.py

@@ -104,7 +104,7 @@ def __init__(
             qf_learning_rate=1e-3,
             policy_weight_decay=0,
             policy_update_method='adam',
-            policy_learning_rate=1e-3,
+            policy_learning_rate=1e-4,
             eval_samples=10000,
             soft_target=True,
             soft_target_tau=0.001,