Using Fitts' Law to Benchmark Assisted Human-Robot Performance