We don't get partial 2-byte and 3-byte UTF8 characters quite correct, fataling instead of returning false because we access past the end of the string.

Instead, bail early if we run off the end of the string without seeing any issues.

Added and executed unit tests.