we may reduce the overhead from heaptrack by writing data out in a binary format. heaptrack_interpret could still use a text format? left to be investigated.
good inspiration: https://en.wikipedia.org/wiki/LEB128
Also an option: http://diamon.org/ctf/