Methods for smoothing/filtering, filter_ functions

[roaldarbol]

Can rely on the signal R package, which again implements the filtering functions from Matlab (I remember having used the filtfilt and butterworth before, they're common in human motor work, so maybe include them too as options). Also see the trajr package for their straightforward implementation of the Savitsky-Golay filter - check to be sure that the default values hold up (https://github.com/JimMcL/trajr/blob/c1d2005f03cc4f17e6cc277371ab26de1bf2bd8d/R/trajectory.R#L372).

Let's try to compile a list of what would be good methods to have:

• filter_rollmean
• filter_rollmedian
• filter_highpass Bandwidth filtering #88
• filter_lowpass
• filter_highpass_fft
• filter_lowpass_fft
• filter_savgol
• filter_kalman

[roaldarbol]

Once data.table supports rolling median on their main branch, move over to that (see Rdatatable/data.table#2778).

[roaldarbol]

Refactor smooth_movement into the above functions, all with the option on_derivatives.