#41 invalid utf-8 workaround

[xbx]

If a document contains the following sequence:
3c 62 3e e3 25 3c 2f 62 3e
"<b>xxx</b>"
where xxx is an invalid utf-8 character sequence, the parser will eat the '<' and say the document is missing the closing element.

The patch works around the situation by making the utf-8 decoder not eating unexepected characters.

[xbx]