incorporate chris's event segmentation model as a preprocessing step for naturalistic stimuli

[andrewheusser]

we could have a find_events function or method that takes an egg and returns a new egg that is segmented into events.

• if k is passed, use that value, otherwise infer it from some function - default: the average within event correlation divided by the average across event correlation + a penalty term
• other model params?