Come up with a new layout of data in S3

[smorrison]

re-organize the data stored in S3 to eliminate duplication. Describe layout schema, how to access entry/revision data, etc.

This will involve keeping external clients in the loop, as at least one partner has direct read access to the bucket.

Take the form of an RFC? Perhaps a github issue/wiki would be enough. This could be structured as documentation for an eventual module for accessing the S3 data.