UTF-8 decoding accepts overlong, stray and truncated sequences #71

DecodeAdvanceUTF8, used by String creation from UTF-8, Bytes.toString and Bytes.getString, does not validate its input. Stray continuation bytes and truncated sequences decode to unrelated code points (89 23 becomes U+0263 and swallows the #), and overlong forms are accepted (C0 AF and E0 80 AF decode to /, which can slip past path checks). Java, C# and Python replace malformed bytes with U+FFFD.