Aerosandbox

- Fixes minor bug that could occur with various functions in `aerosandbox.tools.inspect_tools`, where having `stacklevel` too high could cause the frame to pop off the top of the stack.

- Fixes an AVL bug found by JohannesKimphove.
- Adds version compatibility with NeuralFoil 0.3.0, which offers large speed improvements using caching of neural network parameters.
- Drops support for Python 3.8, to allow better typing and f-strings.

- Verified compatibility with NumPy 2.0.0+; version pin removed.
- Converted formatting to Black, for improved readability.
- Fixed a bunch of minor things using Ruff; removed dead code branches.

- Improvements to AeroSandbox plotting tools (`aerosandbox.tools.pretty_plots`)
- New features in AeroSandbox profiling tools (`aerosandbox.tools.code_benchmarking`)
- In tutorials, added new validation studies for aerodynamics solvers including wind tunnel data

- Fixed a bug in `aerosandbox.numpy.gradient` which resulted in inconsistent answers with `numpy.gradient` (dropped an argument). Added fix for the `period` argument.
- Adds a new tutorial on visualizing flight test data.
- Adds a new tutorial on aircraft noise impact minimization.
- Improved accuracy of transonic modeling with `asb.Airfoil.get_aero_from_neuralfoil()` based on correlation to full-potential and RANS solutions.
- Adds `softmax_scalefree` and `softmin_scalefree` as new surrogate modeling tools. Similar to softmax and softmin, but with a scale dependent on arguments.
- Adds in ability to do `.leading_edge_radius()` calculations in `asb.Airfoil` and `asb.KulfanAirfoil`, which can be used during optimization to prevent sharp leading edges from forming.
- Adds `softplus` and `swish` functions to support neural surrogate modeling.
- Adds preparations for NumPy 2.0 by fixing soon-to-be-deprecated calls.

- Synchronized update: ASB 4.2.3 and NeuralFoil 0.2.0.