I've seen cases where bad encoding strings will result in errors, catching LookupError should solve the problem by falling back onto `chardet` or `utf-8`
Here's one case:
```
textPayload: "Traceback (most recent call last):
File "/opt/conda/lib/python3.7/site-packages/readability/readability.py", line 189, in summary
self._html(True)
File "/opt/conda/lib/python3.7/site-packages/readability/readability.py", line 132, in _html
self.html = self._parse(self.input)
File "/opt/conda/lib/python3.7/site-packages/readability/readability.py", line 141, in _parse
doc, self.encoding = build_doc(input)
File "/opt/conda/lib/python3.7/site-packages/readability/htmls.py", line 17, in build_doc
encoding = get_encoding(page) or 'utf-8'
File "/opt/conda/lib/python3.7/site-packages/readability/encoding.py", line 46, in get_encoding
page.decode(encoding)
LookupError: unknown encoding: utf-8, ie=edge, chrome=1
```
In PYthon 3 .decode() on bytes requires the name of the encoding to be a str type which means we have to convert the extracted encoding before we can use it.
Since get_encoding() is only called when the input is *not* already unicode we need to declare the regexs as byte type so they continue to work in Python 3.
from kennethreitz/requests/utils.py. Also I added some superset
encodings I have found in Chinese pages that are mishandled by
chardet/character declarations.
Éloi Rivard
Yuri Baburov
Dario
Chris Curvey
Martin Thurau
Nathan Breit
Mark Perdomo