write a reusable method to filter out excerpts from low-variance PPA clusters

[rlskoeser]

Adapt from existing notebook exploring the variation of excerpts within clusters. This will be a filter we will want to apply in some cases and not others, so it should be a reusable method.