HTTP compression support

[lovelydinosaur]

HTTP/1.1 compression support should be added at the Stream layer.

• Include an appropriate Accept-Encoding header using the client defaults.
• Inspect and automatically apply the Content-Encoding header on responses.
• Ensure maximum decompression size is not open ended.

I'd suggest starting with just gzip, tho deflate, brotli, and zlib are also options here.

Here's how the .stream accessor might look...

>>> r = cli.get("https://www.example.com")
>>> len(r.body)
128823
>>> r.stream
<GZipStream [128KB unzipped from 100% of 64KB]>
>>> r.stream.raw
<HTTPByteStream [100% of 64KB]>

Related... currently the Connection class returns an AsyncIterByteStream here, tho a custom HTTPByteStream would be neater, even if that's just for the class name / making the providence clear.

We probably initially want just a GZipDecoder implementation, tho we might also choose to implement a GZipEncoder stream implementation for returning gzip responses from the server/application layer...

def compress_response(response):
    if response.headers.get('Content-Encoding') is None:
        status_code = response.status_code
        headers = response.headers.copy_set('Content-Encoding', 'gzip')
        content = GZipEncoder(response.stream)
        return Response(status_code, headers, content=content)
    return response

(In practice it's perhaps often smarter for that case to be handled by service gateway proxies, tho useful to have)