Apply stack deduplication heuristics from BHR frontend to output of bhr_collection.py

[squarewave]

We should apply some of the heuristics of the frontend, notably ignoring nested event loops, ignoring non-mozilla code (except the entry point to it), and ignoring SpiderMonkey internals.

This should dramatically reduce the output size, making the frontend load faster, and preparing us to output daily timeseries data for top hangs.