GIoU for oriented cuboids

Our implementation works only for 3D case. The version for the cuboids aligned with Cartesian axes needs the detection vectors to have at least 8 variables.

In difference to the Mahalanobis metric where the variables are treated uniformly, we need to identify which variables represent position of the cuboids, which encode the orientation and which variables represent sizes of the cuboids. In order to define this, we use the keyword argument ind_pos_w_rz_size of the method .set_association_metric()

...
tracker.set_association_metric('giou-yaw', ind_pos_w_rz_size=(0,1,2,3,6,7,8,9))

The keyword argument ind_pos_w_rz_size defines indices of three position variables, the position of scalar part of the rotation quaternion $w$, position of the z-projection of the rotation quaternion $r_z$ and three size variables. When the argument is omitted, the sequence is ind_pos_w_rz_size=(0,1,2,3,6,-3,-2,-1). Such a choice fits the data layout of the popular nuScenes dataset. See our suggestion for tracking nuScenes annotations.