Improve package download flow

Since #739 has been merged, we now have the ability to redownload and save the package archive (**) after analysis.

It would be ideal if the downloaded/saved archive is the same one that's used for analysis

Some possible ways to achieve this:

1. Downloading the archive first in the worker container, then copying it into the static and dynamic sandboxes. These then do analysis as if on a local package.
2. Have the static analysis sandbox double as the 'package downloader' container, then copy the downloaded files back to the host and into the dynamic analysis container, as well as uploading them to the bucket
3. Have a separate 'downloader' container whose sole job is to download everything, and pass it to the other containers like in option 1.

** One caveat with all this is that Packagist packages hosted on GitLab don't offer direct package archive download URLs. We will need to figure out how to support this case (or possibly fall back to legacy behaviour).

Start by tracing the existing package download, analysis, archive-saving, and upload flow described in the issue and the changes from #739. Define a design that makes the analyzed archive the same one saved and uploaded, while deciding how GitLab-hosted Packagist packages are handled or whether legacy behavior is required.