Write code to integrate object detector that does bounding boxes for both letters and shapes

[EricPedley]

With the new Isaac data, @Dat-Bois has been training yolov8 such that it detects the letters and shapes at once (36 + 9 = 45 classes total for both things). To integrate this, we need to modify the pipeline so that if we get a letter detection on top of a shape detection, it combines them and skips the letter classification step.