updated coco_evaluation.py

[mnorko]

• redefined max_iou value when a match is found to ensure that the match selected is the one with the largest iou value
• match variable can take on the value of 0 which would cause the "if match" statement to evaluate to false when it shouldn't
• defined the eval_id in "true_positives" as the "match" eval_id which stores the correct id for the evaluation match rather than the most recent eval_id
• created a for loop to append false positive eval_ids to ensure that all false positives are accounted for

[andreasveit]

Thanks! I adapted your changes and combined them with the previous pull request #2 .