Add SemanticKITTI LIDAR Segmentation dataset adapter

[tejasstanley]

This PR adds support for loading SemanticKITTI- style LiDAR segmentation datasets in PerceptionMetrics.

The loader SemanticKITTILiDARSegmentationDataset loader builds a dataset index from SemanticKITTI segence folders, reads split definitions and ontology data from the SemanticKITTI YAML config, and integrates the dataset into the exisiting dataset registry under semantickitti_lidar_segmentation.

Changes:

• Added semantickitti.py dataset loader
• Supports SemanticKITTI train, validation and test splits
• Handles both val and valid namiing
• Builds ontology from labels, color_map.
• Verified that samples are collected into a dataframe with points, label, scene, and split columns.

TBD:

• Tutorial supporting MM3D Lidar Segmentation model