Improve the document storage format

Currently, the storage format is loosely defined and can be summarized in:

• page are stored in json files
• sub-pages are stored in directories
• sub-directories named "attachments" contains the attachments for the current page

This is working but:

1. it lacks a formal specification
2. it is not really usable for users accessing the pages externally (i.e., hard to read a flattened json file)