Consider Newer LOWESS Method

[tbpassin]

Confidence intervals for LOWESS models in python has code for a possibly better way to perform LOWESS and to get the confidence limits that the old GF4 algorithm. In particular, it seems to do a better job when the x-axis values are unevenly spaced and even more so when many points pile up on the x-axis.

I would prototype this first with a plugin, and if the improvements hold up, the code would go into GF4 itself.