Pass through bytes in path and querystring

[twm]

Per discussion on #303, arbitrary bytes in the path and query should pass through even if they aren't UTF-8 encoded. The params argument should also accept arbitrary bytes, rather than just ASCII.

The approach here is to use hyperlink.EncodedURL (a.k.a. hyperlink.URL) rather than DecodedURL. The latter has strong UTF-8 opinions. treq continues to accept either type (or bytes or str) as the url argument.

• Closes Can't make request with non-ascii/utf8 query params #303.
• Closes Update "Query Parameters" howto #281.
• Closes Klein #339 #311.

[adiroiban]

Looks good. Only minor comments.
Please check the comments, update as you link and then merge.

Thanks!

[adiroiban]

# A value used to test that it is never encoded or decoded.
# It should be invalid UTF-8 or UTF-32 (at least).
raw_bytes = b'\x00\xff\xfb'

```suggestion
            (b'bytes', ['ascii', raw_bytes]),

[adiroiban]

Should this be hyperlink.EncodedURL ?

I think that examples/query_params.py or inline here, add an example of using hyperlink.EncodedURL together with treq.

from hyperlink import EncodedURL

resp = yield treq.get(EncodedURL('https://httpbin.org/get?raw-value-here'))

[twm]

It very much should not be EncodedURL. IMO that API is dangerous because it doesn't handle percent-encoding for you. If you use it with untrusted input you'll get security injection vulnerabilities! Sometimes you need that lower-level interface, but it's not something I'd want to suggest. If you need it, you'll know.

[twm]

(I think this question speaks to the way that EncodedURL and DecodedURL are confusing names. Particularly since URL is EncodedURL and so seems like the default.)

[twm]

Thank you very much for the review, @adiroiban! I'll merge and release shortly.