OVR-Platform/lidar-validation ? reverse-engineered prompt

Build me a simple Python tool that takes two large point cloud files, one from photogrammetry and one from a LIDAR scan, and aligns the photogrammetry cloud into the LIDAR frame automatically. It should figure out the best match without needing a manual starting guess, use a similarity transform with rotation, translation, and one shared scale, then save the aligned cloud with colors preserved.

I also want it to print a few useful accuracy numbers in the terminal, like the final scale factor, median distance between the clouds, and how many points are within 5 cm and 10 cm. Save the transform details to a JSON file too, so I can reuse it later if needed. If the automatic alignment struggles, let me pass in a rough initial transform file to help it along.

Make it easy to run from the command line with two input files and an output folder, and keep the saved cloud untouched except for the transformed copy.