[Feature Request] Implement full Multi-Camera support

[chordee]

Overview

Currently, the --multi-cams\ flag is available but doesn't change the pipeline's core behavior. Each video file is still processed as a completely separate project with its own database and sparse reconstruction.

Desired Behavior

For a true multi-camera workflow, the pipeline should:

1. Consolidate Inputs: Option to process all videos in a directory as part of a single COLMAP project.
2. Shared Reconstruction: Use a single \database.db\ for all frames extracted from different videos.
3. Camera Identification: Assign different Camera IDs to frames based on their source video file, allowing COLMAP to optimize each camera's intrinsics separately.
4. Unified Output: Generate one sparse model and one \ ransforms.json\ that covers all viewpoints.

Use Cases

• Reconstructing a scene from multiple synchronized camera rigs.
• Combining wide-angle and close-up footage of the same environment.