Deprecate Visvalingam-Whyatt? (Geometry optimization)

[jpt]

after fixing a tricky bug in the AGG-based polygonizer (#36) I'm less convinced this is needed. worth spinning into another library perhaps, but I need re-evaluate different kinds of fonts to determine if it's still useful in three-text

[jpt]

Testing this on more decorative fonts, and complex scripts like Chinese is beginning to change my mind. The right (fixed steps + V-W) is clearly a better result for the same triangle count

[jpt]

well, the answer is no, in any case :)
so I'll close this, but it does still need better documentation (per #13)