The purpose is to provide a class that implements a Node writable stream that can upload data of arbitrary size to Amazon Glacier.
It splits the data into chunks and performs a multi-upload while using the stream to keep RAM usage as low as possible.
Default configuration should send approximatly three requests at the time to Glacier.
More tests Re-upload individual chunks if something goes wrong. Resume failed uploads A mode where it can clean out old options.