KD-Tree implementation with all data in leaves

Question

Before I implement it myself, I would like to ask if there is any KD-Tree implementation that allows to have all actual data in leaves (even if it means duplication of some data for KD-Tree structure itself). I have spent hours googling and only find a more popular KD-Tree version with data shared between inner nodes and leaves.

My goal is to just have rough spatial split for the point cloud and process each leaf individually. I have found also similar question that split point cloud into fixed-size voxels, but I rather want to have my splits similar in resulting number of points (at least within order of magnitude), not just having similar 3D extent.

(Implementation in C++ / Python is preferred. I am just trying not to reinvent a wheel if there is already some well-tested library that I am missing.)