Add RCI Agent

[robert1003]

Implement RCI method proposed in this paper.

Demo trajectories on task config 300 (config_files/300.json) is given here: gdoc

RCI consist of the following step:

• Explicit RCI
  1. Plan: generate initial plan given objective and observation.
  2. Task Grounding (run 1 time)
     a. Critique: criticize the plan given objective and observation.
     b. Improve: improve the plan given the current plan and the critique.
• Implicit RCI
  1. Get next action
  2. State Grounding (run 1 time)
  3. Agent Grounding (run 1 time)