Adding fuzzers for unsafe code.

[adetaylor]

I wanted to make sure that there weren't sneaky ways UTF sequences could mess with the byte-based iteration. No problems found!

[ssokolow]

It's good to see someone is concerned about it. I'm migrating a project from Python's shlex and, being the #![forbid(unsafe_code)] type on what I write and not needing maximum throughput, I've been contemplating offering up a PR to add a paranoid feature flag that switches out the use of unsafe entirely if I can't come up with a way to refactor things to just avoid having to choose.