Improve heuristics when labeling

[phronmophobic]

As of https://github.com/phronmophobic/snowball/commit/94479055403e8819ebc35704c2281540ea448240, the current labeling algorithm adds labels starting with the largest sized dependencies first. If trying to add a label will obstruct a previously added label, then it is skipped.

However, there are cases where long library names will obstruct other similarly large labels and there is enough space to display both names with a little jiggering.

There are trade-offs in runtime complexity, implementation complexity, and effectiveness for different library dependency trees. There are probably some simple heuristics that can be applied that improve on the current algorithm most of the time.