Revert "Fixed multi GPU inference by only applying NMS per worker, not on concatenated results. Add queueing. Also add multiprocessing for star file generation"

This reverts commit bff1185.

Author: mihin-wehi