Add reasoning effort parameter

You can now select the reasoning effort from "low" to "high" for better output of the o3 models