improve documentation, tidy up implementations and be more consistent about usage
find symmetric x:
currently the way it does windowing is not nice and can lead windows in the wings of a distribution being preferred. It's not clear to me what we can do about this. pick windows based on y scale?
currently relies on a regular grid because it directly compares points. Move to RMS?
good at rejecting noise
peak: bad noise rejection
fft shift
less good without regular grid
needs whole function to be sampled (e.g. won't do a good job of lorentzian centred around zero)
good at rejecting noise
sampling - need to know all other parmeters otherwise excellent
improve documentation, tidy up implementations and be more consistent about usage