Smoothing weights should be inverse stdev

[kdkorthauer]

The weights provided to the smoothing function should be the inverse standard deviation of the methylation difference coefficient, since we want to place more weight on observations with less variation.