Using gpt-3.5-turbo to format the result from Code Execution to reduce cost.

[myfunc]

Use case:
I use the default gpt-4 model, and when User Proxy executes the code, it sends the entire result to chat. Sometimes this can be more than 6k tokens with useless information.
Performing a simple task can cost a few dollars because of this.

To analyze the long result I see a solution:

• Use gpt-3.5-turbo to analyze the output to the console and make a short summary with key points, then pass it to the Coder agent.
  This would reduce the cost a lot.

[sonichi]

Clever! You may be interested in #131
@kevin666aa for awareness.