Refactor extraction module

[RyanDoesMath]

This pull request makes a few important changes to the extraction module.

1. The detection of objects is now isolated to a single function called detect_objects_using_tiling.
2. Functions have been added that detect all the objects on both the intraoperative and preoperative/postoperative sides of the image, and returns a dictionary of detections.
3. Functions have been added that assign meaning to the detections dictionary.
4. Adds a new module for point registration. When performing only meaning-assignment to detections, it is not necessary to do whole-image registration.

The functions that did both at the same time have been left alone, since they get slightly different results by detecting on the post-homography images.