Use NeuralFoil instead of Xfoil

[1-Bart-1]

NeuralFoil is pretty cool and a lot faster than Xfoil. Xfoil is problematic because of the @eval @everywhere macros.
https://github.com/peterdsharpe/NeuralFoil

• removes the need for parallel processing, because it is so much faster
• doesn't give NaN's which are bad for nonlinear solvers

[ufechner7]

This would require to add Python as dependency. Can be done, but would it be worth it?

[1-Bart-1]

Why not? Everyone already has downloaded python3, might as well use it.

[ufechner7]

Why not? Because if we want to use this package in KiteModels it should be as small as possible. KitePower needs to run their simulations in a docker image, and the model size should not exceed 1GB. Adding NeuralFail here would significantly increase the package size.

We might have to split this package into two packages if we want to add more dependencies.

[1-Bart-1]

We could decide to move core functionality to a new package VortexStepMethodBase.jl in the future