Rethink the use of the Hample filter to smooth step times in CAD calculation.

[AKuederle]

I only quickly took screenshots of the two cases and did not investigate deeper, but basically, I was looking at the Hample filter used for smoothing in the CAD estimation. I had a case, where the input ICs where only shifted by 1 sample, but the output was significantly different. (input in the screenshopts -> ICS, output -> step_time_per_sec_smooth)

The intersting thing is, that in one of the cases the values are basically all the same in the output.
What I think what is happening here, as the Hample filter is calculating the SD over 3 samples and then comparing with the remaining samples, things become difficult, when the SD becomes very low, as this basically means that any other value is considered an outlier and smoothed
So if you have two Cadence values that are identical, it might be that everything collapses, because the STD is to low.
Similar things should also happen for SL as it uses the same type of filter.
I have not verified, if the same thing also happens in the Matlab implementation, but I assume it does.