Preprocess

The script performs preprocessing on pointclouds data (.pcd files) 
 and their respective asc files (labels) and saves the processed data (tensors) 
 as pickle files.


It includes the following steps:

1. Loading PCD files and the corresponding label files (ASC files).
2. Normalizing the point cloud data by translating it to the origin and scaling.
3. Downsampling the point cloud data using voxel downsampling.
4. Estimating the normals for the downsampled points.
5. Assigning labels to the downsampled points based on the K-nearest neighbors in the original point cloud.
6. Converting the point cloud data (coordinates, colors, normals, and labels) into NumPy arrays.
7. Saving the preprocessed data as PLY files.


Usage:
Set the appropriate configuration values in `config.py` 
 and run this script directly to preprocess all PCD files in the
 specified raw_file_dir.