update deprecated code (linear_assignment and np's type)

[tannd-ds]

• the linear_assignment from scikit-learn is deprecated and has been removed, instead, now we use linear_sum_assignment from scipy.optimize. The two functions have slightly different return types and I did adjust the code so that it worked as it was.
• Many type casting in the project are deprecated (like np.int is no longer used, instead we use int or specify the precision like np.int32).

I've already tested it and now it runs perfectly without any errors.

[FlyingPotato437]

I agree, should be replacing the cost matrix functino with this:
cost_matrix = distance_metric(tracks, detections, track_indices, detection_indices)
cost_matrix[cost_matrix > max_distance] = max_distance + 1e-5

Use linear_sum_assignment from scipy.optimize

row_indices, col_indices = linear_sum_assignment(cost_matrix)
indices = list(zip(row_indices, col_indices))