Potentially try / compare to a Gaussian Mixture Model

[TomDonoghue]

Idea:
Why not use a Gaussian Mixture Model to select the number of peaks. This could be done by plotting the BIC scores for varying N and the GMM with the lowest BIC is the one you would select.

Notes:
Early versions of FOOOF did try a GMM. It was much slower, and at the time didn't work that well, but this could well have been due to issues isolating the aperiodic component. This is potentially an idea worth revisiting / comparing to.

This idea brought up here:
fooof-tools/fooof#114