Feature: Simplify Custom Chain Execution with .chain Files (Chain of Thoughts format)

[PatrickDiallo23]

Introduce a feature that processes .chain files to create and manage custom chains of request execution. Each step in the chain corresponds to a specific prompt (or task), and steps are organized into (sub)folders for modular management. Also, those steps can be automatically executed or done with user confirmation (ask before moving to the next step). Changes to steps will trigger rollbacks and re-execution of subsequent steps.

Example Workflow:

1. Input .chain file:

Implement an application that collects user data and saves it in [[file]].csv.  (01_step)
---  
Cover it with tests.  (02_step)
---  
Add REST endpoint for [[file]].csv.  (03_step)

Note: LLM can be asked to split a bigger task into these steps.
2. System generates:

chain/01_step/01_step_prompt.txt
chain/02_step/02_step_prompt.txt
chain/03_step/03_step_prompt.txt

Note: Optionally, a batch file to execute steps sequentially can be created as well.
3. Execution proceeds step by step (with or without user approval), with rollback if step modifications occur.

A new directive (#chain-confirmations: on/off) can be implemented in .chain files to toggle between fully automated execution and user-confirmed execution. This allows users to customize how the chain steps are processed.