Refactor AgentState for future Task requirements

[TanGentleman]

Changes:

• Move all task state to the Task object.

For each new set of Actions / Task Type:
Define side effects (mutation of AgentState):

• At TaskManager node to set Task Status to IN_PROGRESS
• At Action node to continue flow after ActionResult is obtained (define by case)

Questionable (is handoff logic needed?):

• At TaskManager node: Status.DONE or Status.FAILED
• At TaskManager node: When stash action is queued

Code changes:

• For each TaskType — init logic in Task Manager node.
• Add logic for each ActionResult case after action execution
• Define Stash/HITL logic (by Task)

[TanGentleman]

Been grinding this out. I've made great progress!

Moving forward, I want to modularize and clearly define the state mutations for new Action/Task types.

I'll try making better use of Python's case syntax, but so far abstracting the task_state to the Task object has been really helpful!