Parallelise agents

[ElliottDyson]

Agents currently occur sequentially.

Try and develop an optional alternative method that runs a different LLM instance for each character for real-time suitability to instances such as interactable RPGs.

Have each agent "wait" until a timestep has been completed, this ensures alignment between agents to ensure agents do not become out of sync.

During interactions, we should have agents be paused whilst other agents are attempting to interact. This allows the agents to not interrupt each other and therefore become confused.