Complete Agent class

[AidanTilgner]

The Agent class should be completed. There are a few pieces of functionality that need to be considered here:

• Action Stepping: the agent should be able to step through every action in the Plan of Action
• Post-Action Decision: the assistant should make a decision after the action is taken, between a few different items: add_to_context, mark_complete, mark_failed, and possibly more. The Agent should be able to pick multiple actions to execute in order
• Clarification: the Agent should be capable of sending a message to the user asking for clarification on a certain item, then when a response is received it will process the response. The user should be able to specify in a message whether they're talking to the assistant, or a specific Agent by adding an agent property to the GlobalChannelMessage object.
• Context-Based Plan Revision: the Agent should be able to get a revised version of it's plan of action based on new context or other information gathered
• Assignment Updates: if a user sends a message, and the plan of action seems to overlap with what an existing agent's current task, then its plan of action should be updated to include the new information.

There may be more things that we decide to add in the future of course, but this list of functionality should get us started in terms of making Agents useful.