Subclass bytes to support the Buffer API

[dstufft]

[jamesls]

Not sure what to make of the travis failures, any ideas?

[dstufft]

I'm not sure, I'm poking at it but it appears the event handler that is being registered by TransferManager isn't getting called at all.

[tipabu]

FWIW, I think the failure had to do with CallbackEnablingBody always being a False-y value, meaning that when we'd prepare the request, we'd pass a None body to requests. See https://github.com/boto/botocore/blob/@{2017-07-05}/botocore/vendored/requests/models.py#L455 -- since data is False-y, body stays None.

Even with modern botocore (which has its own request-preparer, no longer vendors requests, etc.) the body == b'' check in AWSRequestPreparer._prepare_body (see https://github.com/boto/botocore/blob/1.12.26/botocore/awsrequest.py#L411-L412, for example) would cause the same behavior.

Regardless, pretty sure this got addressed by #91, when CallbackEnablingBody started subclassing RawIOBase -- I think this PR can be closed.