Move processed tarballs to archive bucket

[bedroge]

In order to reduce the number of GitHub API calls and increase the performance of the script, we should move processes tarballs to another (archive) bucket. This will prevent the script from reprocessing them for every run.

[trz42]

The new method actually looks for metadata files and these could be moved away. So one could do the following:

1. Upload metadata (or later .task) file to a top-level directory new and tarballs to a top-level directory tarballs (the bot is already prepared, ie, all is needed is to adjust the configuration a little bit (settings metadata_prefix and artefact_prefix)
2. Scan directories new and in_progress for metadata files and process them as usual (for all in new we have to move them to in_progress as soon as we see them)
3. Once a staging PR is approved/rejected, move the metadata file to approved or rejected (or just to done).

2 and 3 require some additional coding. It shouldn't be too difficult though.