Get object sizes based on S3's ListObjects output (just with scan_object_sizes() to start with)

[poodlewars]

Monday: 8560764974

Limited to just scan_object_sizes to start with, to show the idea.

Calculate object sizes using functionality in the storage backend itself when possible. This PR calculates compressed sizes on S3 based on the ListObjectsV2 output, so we don't need to read all the keys.

For storages where we haven't implemented anything fancy, just read all the (compressed) keys and check the compressed size on their header.

Remaining work:

• Use this approach for scan_object_sizes_by_stream
• Make sure we have testing against all backends especially NFS
• Implement scan_object_sizes_for_stream
• Native size calculations for LMDB and Azure
• Native size calculation for library size (LMDB has an API to do this "in one")
• AdminTools Python API on the v2 API, including search by regex. Docs page to explain the output.