`dt` not passed through properly in `jerk_sliding`

[pavelkomarov]

I've crafted a test case for it as part of #117, but even before my changes, it appears to just want to find a constant derivative. Choosing different gamma weighting does not fix it.

[pavelkomarov]

I've decreased the window size cutoff from 1000 to 100 and shortened the kernel by a factor of 10 kernel = np.hstack((np.arange(1, 21)/20, np.ones(60), (np.arange(0, 21)/20)[::-1])), and now the function fits a bit better, but still not well (gamma=1 here):

I'm not convinced this bug is anything more than the solution not being very good when there are too many points. Here is the same test with 100 points, which is just short enough that the method passes through to the standard jerk function instead:

Prettymuch identical.

[pavelkomarov]

Oh, I wasn't passing through dt properly for this case!