Redistribute responsibilities between workspace and strategy cache #70
Description
Right now, each workspace has to carry an actions vector, which might not even be used depending on what strategy_cache is used (GivenStrategyCache). If strategy_cache carries the actions vector, we can prevent this. Furthermore, we already dispatch state_action_bellman depending on the strategy_cache type and could possible make it more generic, such that it is shared among all model types and state_bellman is the workspace dependent function.