2D polar extrapolation is not realistic in lift

[jellepoland]

Currently, the 2D polar (for V3 kite) extrapolates as follows:

This is an artifact of the V3 LEI airfoil input data only stretching until 24.5 °, and it extrapolates in a straight line afterwards.
This is clearly wrong, given that a flat-plate airfoil shows approximately this behavior (source: https://digital.library.unt.edu/ark:/67531/metadc57443/)

I suggest incorporating a flat-plate model for lift, drag, and moments; that becomes active beyond the range of the input 2D polars.
Another road would be to switch to NeuralFoil if possible in this extended range.

[1-Bart-1]

I would prefer to use neuralfoil for everything, as a more general solution that is easier to maintain. We can even just use the neuralfoil weights so we don't need a python dependency.