1. Now we are calling into requests library and the default behavior is we de-compress the data if there is a "transfer-encoding' header or "content-encoding" header.
2. In storage SDK, if user wants to download a big file from blob, storage sdk splits it into multiple 4m chunks.

Here is the broken case:
User uploads a big (>4m) gzipped file into blob and sets the content-encoding to gzip then he/she calls storage sdk to download the file.

Storage SDK splits it into multiple requests, for the non-first request, azure-core recognizes the "content-encoding" header and tries to decompress the stream -- fail