Simplified and faster encoding.py & doc building

[adbar]

• code cleaning: regexes + LookupError unnecessary as fix_charset() returns a default value
• faster encoding detection with cchardet, made optional in encoding.py and setup.py

[adbar]

The conflict comes from setup.py, please review it

[adbar]

• added bypass of doc_build

[buriy]

This option doesn't set self.encoding which was set in another branch.
Users might be relying on it.
So please add
self.endoding = 'utf-8'
to match the expectations in the best way and we'll publish it then.

[buriy]

Also, you have no deepcopy as we have discussed.
Any thoughts on this?
._parse method could be run several times, devastating the document after the first iteration and thus reducing the found subtree quality.
See

python-readability/readability/readability.py

Line 219 in 4dcd6f9

self._html(True)

and

python-readability/readability/readability.py

Line 197 in 4dcd6f9

return shorten_title(self._html(True))

and etc.

[adbar]

I added the encoding attribute, for the deepcopy issue I'm not sure what to do, passing already existing trees is experimental anyway

[adbar]

@buriy Could you please accept or amend the PR? All tests pass and we could keep the deepcopy improvement for later.