Fix prescan bug: open end tag and alpha ('</' + alpha)

```
#data
</xxx<meta charset=iso-8859-2>
#encoding
None
```

When reading '</x', A prescan parser should jump to space (' ') after 'meta', ignoring '<'.

While `html5lib._inputstream.EncodingParser` returns: 'iso-8859-2'.