Refactor DocumentID

[amorton]

Following from #2378

• improve the management doc the DocumentId Type and the encoding values for the db tuple
• currently the code is not symetric, we can create a EjsonDocumentId from json but not from reading from the db. This is because the Ejson DB encoding type is set to string, so the db compares them as strings, but this means we dont know how to rebuild it from reading from the DB :( This is not ideal because we have code that rebuilds from the other types.
• Improve the layout and quality for DocumentId, it's not bad code just needs some polish