What i would love is to take a "team" of 10 cars with a IA pilot to do the dirty job in testing.
They would go all on indentical setup except on one parameter.
For exemple the ride higth. The ten cars have a value in a range I choose.
if it is 50 to 95 mm then the cars would have these values : 50 55 60 65 70 75 80 85 90 95.

They all start at the same place and time (with collision off) and then i see corner after corner witch behave best.

This way I would be able to try many options for a tenth of time. And then "finish" the setup myself the usual way.

The LFS AI drivers are not perfect robots, so such as test would tell you very little. They make mistakes, they learn from their (differing) experiences, so it is therefore feasible and likely that the optimum setup'd car would not necessarily be fastest.

The best way to check a setup is to drive consistent laps yourself. Setups are very personal, so one that 'works' for the AI might feel horrid to you. And if you are not consistent enough to do the setup checking yourself, then I suspect that the subtle improvements these changes bring would not help you.

However, you can tell the AI to use a setup. First make a setup, and select it in the race config screen. Add an AI. It will have that setup, and use it no matter what you do now (other than remove it). Then make the next setup, and add an AI driver...

When you have 10 setups on 10 different AI, start the race, and let them get on with it. It would help if the AI's all have 300 - 400 laps experience with the same setup, and produce VERY similar laptimes prior to giving them different setups.