Currently we assume content-length is the full size of the download file, but it's actually the length of the current response body.

I.e, if we request 100-, the content length of this response is 100 bytes smaller than the full size of the file.

We need to fix the function to chunk the file and adjust the unit tests, since it's based on a incorrect assumption.

Optionally it will be good to see if other code in download system has the same issue.