[WIP] Upgrade tensor-based RGB-D Reconstruction System

[theNded]

New modules

• [x] --fragment_registration: point cloud odometry and global registration with tensor ICP and correspondences from feature matching
• [x] --fusion: final reconstruction after joint optimization

Improvements

• [x] pose graph wrapper's visualization module

TODO (minor)

• [ ] Policies for point cloud loop closure detection: [exhaustive, windowed, random]
• [ ] RGB-D loop closure from features (temporarily cv2's ORB feature)
• [ ] Multi-scale ICP and refined pose graph optimization
• [ ] CPU/CUDA/Legacy switcher

Future work

• [ ] Fast global registration with CUDA
• [ ] Mean reduction for voxel downsample
• [ ] GUI (look into COLMAP)
• [ ] Interface for learned loop closure detection (external learned image/point cloud descriptors)

---

This change is 

[update-docs[bot]]

Thanks for submitting this pull request! The maintainers of this repository would appreciate if you could update the CHANGELOG.md based on your changes.