Keep track of detection index

[abaybektursun]

Addresses #195 and unanswered question in #5:
Currently, after updating ByteTrack object, some of the bounding boxes will be dropped, and it's not possible to figure out the original bounding box index form the tracker. For example

boxes, scores, classes, .... = some_detection_model.inference(frame)
boxes_scores = np.concatenate([boxes, scores.reshape(-1,1), ], axis=1)
# In this step len(online_targets) <= len(boxes)
online_targets = tracker.update(boxes_scores, [frame.shape[0], frame.shape[1]], [frame.shape[0], frame.shape[1]])
# Since we are missing some of the boxes, there is no way figuring out relevant information of the boxes like class.
# I did bbox matching method to figure out the original index, but that's inefficient (N^2) 

With the solution in this PR you can track index directly using tracker.update(..., track_det_idx=True):

boxes, scores, classes, .... = some_detection_model.inference(image)
det_idxs_orig = np.array(list(range(post_boxes.shape[0]))).reshape(-1,1)
boxes_scores = np.concatenate([boxes, scores.reshape(-1,1), det_idxs_orig], axis=1)
online_targets = tracker.update(boxes_scores, [frame.shape[0], frame.shape[1]], [frame.shape[0], frame.shape[1]], track_det_idx=True)

for track_i, an_online_target in enumerate(online_targets):
            track_xyxy = utils.tlwh_to_xyxy(an_online_target.tlwh)
            det_i = an_online_target.det_idx
            tracked_obj_class = classes[int(det_i)]

[abaybektursun]

@ifzhang?