unicode/utf8: FullRune returns false for a known invalid byte

Any UTF-8 sequence that  starts with 0xc0 or 0c1 is ill-formed and DecodeRune always returns RuneError, 1 for such a byte. However, FullRune of []byte("\xc0") returns false, even though the spec says it will return true for known invalid bytes.

I can look into fixing this as part of a considering a table-based utf-8 decoder. A fix would be trivial if we adopt this algorithm.
