`calculate_rewards()` is defined the wrong place

[jbirkesteen]

The way we define the calculate_rewards() methods in Player() is actually specific to direct policy gradient methods.
Therefore, the definition in Player() should be an empty placeholder, while the current definition is moved to the appropriate DirectPolicyAgent() parent class.

[jbirkesteen]

When this is fixed, we should remove the definition of calculate_rewards() from the TD-agent, so it just inherits the pass from Player().

[jbirkesteen]

When this is fixed, we should remove the definition of calculate_rewards() from the TD-agent, so it just inherits the pass from Player().

TDAgent still needs to pass it, since it inherits from DirectPolicyAgent and not Player.