feat(core): Butterworth filter

WHY IS THIS CHANGE NEEDED?:

• as an alternate way to remove long-term trends in time series -- applying a low-pass filter.

HOW DOES THE CHANGE SOLVE THE PROBLEM?:

• general wrapper for scipy.signal.butter in case the process is useful for other things
• renamed parameters from scipy.signal.butter to make them more informative
• wrap filter design (butter) and filter application (sosfiltfilt) in butterfilter -- they are used together anyway. and this makes pandas operations in run() easier to read
• subtract mean to correct initial time points
• applied filter twice to eliminate phase shift
• grafted the original column names onto the output to preserve time units

WHAT SIDE EFFECTS DOES THIS CHANGE HAVE?:

• chucks out time series that have NaNs

EVIDENCE THAT COMMIT WORKS:

• tested with the examples in https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.butter.html and modified parameters to match -- output identical to shown
• tested with real data

REFERENCES:

• Issue #25 (closed)