Copy the compressed bag to a working storage bucket before we start to unpack it

[alexwlchan]

Right now we unpack the bag directly from the user uploads bucket. We don't actually read the file in one go – we read it in pieces which get stitched together, to avoid dropping the S3 connection.

Theoretically the bag could change under our feet while this happens; it might be expedient to copy the compressed bag into temporary storage before we start unpacking it.