Return the State of the environment that results after Agent a
executes the Action in State s. This reflects the state transformer
function T:R^Ac -> 2^E for a specific action in Ac, with two
caveats. First, if the action is nondeterministic, this method
selects one of the resulting states. Second, if history
is important, we assume that the state s contains all necessary
history information. Note that in a multiagent environment, it is
also important to know which agent is executing the action.
DESIGN NOTE: The choice to divide the state transformer into
separate methods for each action is essentially one of modularity.
We suggest that having each action "know" what its effects are
is more natural than having the environment (or each state)
know how states are changed by actions.