HTMLParser fails if a multi-byte character falls on a 4K boundary

[hodgestar]

If one does:

text = u'a' * ((4 * 1024) - 1) + u'\xe6'
events = list(HTMLParser(BytesIO(text.encode('utf-8')),
                                 encoding='utf-8'))

it produces a truncated-input error because the multi-byte character crosses the boundary of a read from the input file.

[hodgestar]

Fixed in r1189.