Re-implement the index-to-storage method#43

Right now Pentagon is implemented in a way where for every index (eg. primary, unique, index) we create a duplicate entry. It's not optimal especially when we have more data. For instance, in the case where we want to fetch items only after a certain `createdAt` timestamp, currently we would need to fetch all the items, and then filter by `createdAt`. Instead, what we want is to be able to index by `createdAt`, only list the keys, and only after that fetch the actual items. [Denodata](https://github.com/anywhichway/denodata) has a great implementation of how this should basically look like.