Feature Request: expose a target_update method for DQNAgent

[billh0420]

Currently, DQNAgent updates the target network periodically.

Would it make sense to provide a target_update method in DQNAgent to update the target network?

Reasons:

1. I find it confusing to determine how often and how frequently the target network is being updated.
2. If I could call the target_update method explicitly in my training loop, then I could easily determine how often and how frequently the target network is being updated.
3. When the training loop ends, I can insure that the target network is updated at the end (before the agent is being saved).