Added chunk_size to Response.iter_bytes() and Response.aiter_bytes() (#394)

[cdeler]

Working on #394 , I added the chunk_size argument to Response.iter_bytes() and Response.aiter_bytes() wrapping these iterators by drain_by_chunks() and async_drain_by_chunks() (I didn't use decorators to make debugging easier). The same way might be applied to (a)iter_text

But I still don't know what the expected behaviour of (a)iter_line and (a)iter_raw (so therefore I omit them in the PR)

For now the code might look like:

>>> import httpx
>>> response = httpx.Response(200, content=b"1234567890")
>>> for it in response.iter_bytes(3):
>>>    print(it.decode())
123
456
789
0
>>> # it's done for requests compatibility: 
>>> # https://github.com/encode/httpx/issues/394#issuecomment-598820965
>>> for it in response.iter_bytes(None):
>>>     print(it.decode())
1234567890

[tomchristie]

This one's a bit fiddly, so I'm going to defer on reviewing it for a little bit.

[tomchristie]

So, I'd been planning for us to take an approach more along these lines... #1277

That adds a decoder class,ByteChunker which has the same kind of interface as all the other decoders,
and which doesn't involve having separate sync + async implementations.

I've not quite finished that work off, since it still needs:

• Using ByteChunker within iter_bytes.
• Adding a corresponding TextChunker class.
• Using TextChunker within iter_text.

[cdeler]

@tomchristie
It's better to close this PR?

[tomchristie]

@cdeler - I think so, yup - thanks for having taken a crack at it.