build, upload, and host graphviz output

[kalexmills]

To avoid more manual effort, VetBot should build graphviz visualizations of suspicious callgraphs, upload them somewhere (probably Google's version of an S3 Bucket) and embed the picture into the uploaded file in the issue.

Originally I wanted to keep the images in the same repository as the issues, but I don't like that so much since it uses at least one more API request (and, anecdotally, we're already hitting the rate limit between VetBot and TrackBot).

This is a nice-to-have feature, which should probably be done anyway to make the UX better for our crowdsourcers.