Add option to use COON while code generation to reduce the tokens to 30-50%

[AffanShaikhsurab]

We can add the ability to use the COON format for the currently supported frameworks (Flutter and React). This way, the AI will generate a compressed code version of the code , and we can decompress it before writing to the file saving around 30–50% in token usage.