Garuda

Populated the following properties in `ConfusionMatrix` class
* true_positives
* predicted_positives
* false_positives
* actual_positives
* false_negatives
* precision
* recall
* f1_score
* summary

* `from garuda.ops import obb_iou` function is implemented as numpy variant of `probiou` in `ultralytics`. However, `obb_iou` works more like the `iou` function of `supervision` i.e. given `n` ground truth boxes and `m` predicted boxes, it will return `nxm` IoU matrix, unlike `ultralytics` probiou which takes `n` ground truth boxes, `n` predicted boxes and returns their one-to-one IoU vector of size `n`.
* Added `from garuda.od import ConfusionMatrix` pipeline for computing confusion matrix for OBB detection. This is heavily inspired by ConfusionMatrix from `supervision`.

* Added label-studio CSV to YOLO OBB conversion functions
* Added bulk operations
* Save all labels in YOLO OBB format to a directory from a label-studio CSV
* Read all labels in YOLO OBB, convert them to YOLO AA, and save them to a new directory
* Added testing for README snippets

Initial release
* Added basic functionality and populated README with useful information.