Add low pass filter on velocity signal for idle time scoring function

[shreetej-reddy]

We should probably low pass the velocity signal in the idle time scoring function to reduce noisiness/spurious idle measurements.

We should also take a pass and see where else filtering might be a good idea

[trudramukerji14]

This is a good point! I think this use of np.diff comes up in multiple scoring functions.

One thought with noise in idle-times is that may cause us to miss some idle-times because the noise may cause some states to appear to be moving between the time-stamps when they are not. This may be partially addressed by the threshold? But I agree, a filter like a convolution could be helpful in smoothing the signal.