I'm suggesting you use an unstructured mesh. The benefit is that it's easily made to follow the STL surface. Boundary layers are easily made on the wing surfaces, so you don't lose accuracy from that. I understand that there will be more elements in an unstructured mesh than a structured mesh. That means the solver will run for longer and maybe a computer with more resources is required. But the time to make the mesh for each new wing shape will be much less than making a structured mesh on a solid model.

A huge amount of time will be spent making a solid model of each wing from the STL data and approximations to the shape will be introduced in the process. I don't think the benefits of a structured mesh can justify all the time spent making one for each new wing shape.