Documenting byte streams

[lovelydinosaur]

Our documentation doesn't currently cover the byte stream interface.

We should document why we have a byte stream abstraction, and link to it from the Content Types, section.
For example we need to document the .encode() -> [Stream, str] interface for returning the content and content type.

Examples look a bit like this...

>>> r cli.get('https://www.example.com')
>>> print(r.stream)
<ByteStream [100% of 64kB]>

# Response with Content-Length
>>> with cli.stream('GET', 'https://www.example.com') as r:
>>>     print(r.stream)
<ByteStream [0% of 64kB]>

# Response with `Transfer-Encoding: chunked`
>>> with cli.stream('GET', 'https://www.example.com') as r
>>>     print(r.stream)
<ByteStream [0% of ???]>

# Response with `Transfer-Encoding: chunked`
>>> with cli.stream('GET', 'https://www.example.com') as r
>>>     r.read(30*1024)
>>>     print(r.stream)
<ByteStream [???% of ???]>

# Response with `Transfer-Encoding: chunked`
>>> with cli.stream('GET', 'https://www.example.com') as r
>>>     r.read()
>>>     print(r.stream)
<ByteStream [100% of 64kB]>

See related #30 which has an interesting note on HTTPByteStream as a class name. This would also be useful because the stream under-run/overrun in that case could raise a slightly more explicit error than the generic ValueError case.