Moving timeout and max steps into `PaperQAEnvironment` itself

[jamesbraza]

Currently as of https://github.com/Future-House/paper-qa/tree/v5.0.7:

• Timeouts are managed by rollouts in agents.main
• There is not a max steps, unless using ldp

It would be nice if the PaperQAEnvironment itself directly supported timeouts and max steps. That way we can have:

• A simpler agents.main
• More information (e.g. step count, current runtime) directly in env.state