Autocast: single-agent + fully observable -> multi-agent, working only under specific hypotheses.

[nhuet]

When a multi-agent solver deals with a single agent fully observable domain, it only works if the state is also a dictionary of agent's state, which is not enforced by the current API.