Faster version of the computation of radius neighbors for 3D Point Cloud

[humanpose1]

because the computation of the radius neighbors on cpu is very slow, I decided to modify it a little bit.
The code uses the header Nanoflann to build the kdtree, unfortunately, it only supports dimension 3.

What do you think about it ?

[rusty1s]

Wow, this looks impressing. Thank you! Do you have some benchmarks to justify your speed-up claims?

[humanpose1]

Thank you for your answer !
Here is some plots to show that nanoflann is faster than scipy.

Also the output of my function is a torch tensor whereas scipy gives a numpy array so you have to convert it into a torch tensor.
The code to perform the experiments can be found Here

[github-actions]

This pull request had no activity for 6 months. It will be closed in 2 weeks unless there is some new activity.