[coverage] tokenizer.c: valid_utf8

[valid_utf8](http://droettboom.com/cpython-coverage/llvm-coverage/coverage/home/mdboom/Work/builds/cpython/Parser/tokenizer.c.html#L493) seems to never be fed invalid UTF-8 by the test suite.