FourierRadon2D and FourierRadon3D

[mrava87]

This PR introduces 2 new operators, namely FourierRadon2D and FourierRadon3D that perform linear/parabolic Radon transform in the frequency domain.

Both of them have 3 backends:

• numpy: can run with both numpy and cupy arrays, suboptimal performance
• numba: optimal performance for numpy arrays
• cuda: optimal performance for cupy arrays (uses a numba-cuda custom kernel for part of the operations in matvec/rmatvec

To test the parabolic transform in 3D, a new utility function pylops.utils.seismicevents.parabolic3d is also added.

The new operators are showcased in plot_fourierradon.py. Moreover, two notebook have been added at https://github.com/PyLops/pylops_notebooks/blob/master/developement/Radon-Fourier.ipynb and https://github.com/PyLops/pylops_notebooks/blob/master/developement-cupy/Radon-Fourier.ipynb

[cako]

Left some comments and made a few changes. Unfortunately I'm having an error with the numba engine (if I remove the @jit it works) though. Will try to dig a little deeper, but for now, after the changes, will be good to go.

[mrava87]

@cako thanks! I replied to some of your comments, and I am going to act immediately on others.

I am a bit confused about two things:

• Unfortunately I'm having an error with the numba engine (if I remove the @jit it works) though.: you think this is problem with your machine? And you refer to numba.jit or cuda.jit or both? I tried to re-run those examples on my MacBook pro, Linux workstation, and Colab and I am able to run both numba and cuda codes 🙈
• I saw you made some changes in the `signalprocessing/init.py file, basically the order of some of the imports changed... and I think this is the cause of all tests failing as I get the same issue locally. But if I bring back the original ordering of imports things get fixed in my local machine and also in the CI. I am curious to know if this new order working for you? And do you get why a different importing order can create this problem (I assume it is probably because in the new order we import a module that requires another module imported later, but I am not so sure myself.. didn't do any test so far)

[cako]

@mrava87 Sorry, you can revert the order of imports back to the original. My stupid ruff reorders the imports automatically and I didnt catch it in the commit. My bad!

I'm not sure what the issue is, I tried different versions of Numba as well. All CPU. If it works on yours, I'm not keen on holding back the PR unless I know what the problem is. I can file a bug if it still doesn't work later.