Adding tests of invalid utf-8

[Sytten]

Hi!
I saw the recent changes to the UTF-8 parsing (on HN) and while speed is always appreciated, there is a whole class of bugs related to bad parsing of UTF-8 strings (https://en.m.wikipedia.org/wiki/UTF-8#Invalid_sequences_and_error_handling, http://www.unicode.org/reports/tr36/#Canonical_Represenation). Given that, I am thinking that code path should be fuzzed/tested a lot more to make sure the clever tricks correctly handle intentionally malformed input. I am happy to contribute some tests, just opening the issue to get some feedback.
Thanks