- Removed redundant http head requests that were used to query filesizes, even after the filesize was known already.
- Added `download_feedback_multi` that downloads one file using multiple connections. - The file is preallocated and zero-filled and then written to in parallel at different offsets.