Support for LiDAR segmentation gui

[tejasstanley]

This PR adds GUI support for LiDAR semantic segmentaion.
Current datasets supported: SemanticKITTI

The LiDAR segmentation task includes a dataset viewer, inference tab, and evaluator tab. The dataset viewer loads SemanticKITTI frames, renders point clouds with Plotly, and supports coloring by semantic labels or intensity. It also includes frame selection, point-size control, max-point subsampling, top-view reset, hover information, and a class table.

The inference tab wires the loaded LiDAR segmentation model to uploaded .bin point cloud files and visualizes both predicted semantic classes and intensity. The evaluator tab uses the PerceptionMetrics model.eval() flow for SemanticKITTI and includes live progress updates during evaluation.

Key additions:

• Added SemanticKITTI dataset loading in the LiDAR segmentation GUI.
• Added Plotly-based 3D point cloud visualization.
• Added semantic-label and intensity-based coloring.
• Added intensity colorbar for low-to-high intensity values.
• Added point hover information with coordinates and class names where available.
• Added LiDAR inference support for uploaded .bin point clouds.
• Added LiDAR evaluator support with progress callbacks.
• Added sidebar inputs for SemanticKITTI dataset paths, config YAML, and model files.