Update the data structure and what is signed by witness proofs

[swcurran]

The data model for the did-witness.json file is not correct in the spec. right now. Will correct it.

In addition it was agreed that we would:

• make the versionId the object that gets signed -- not the entry.
• The embedded proof is an array of proofs from witnesses signing the versionId
  • This matches the Data Integrity model of the proof being added to a data structure it is signing.
  • This also allows the entire set of witness proofs to be verified using only the data in the did-witness.json file.
  • This would mean that there is no way to know if the witness actually verified a new DID Entry before approving it, or did they just blindly signed the versionId value. Is that OK?
• When deleting proofs, if a versionId has no proofs associated with it, it can be deleted from the file.

Please confirm that this is right @andrewwhitehead @PatStLouis @brianorwhatever and others. @brianorwhatever (or someone) -- please let me know the data model for the did-witness.json file.

[brianorwhatever]

Yeah, that looks correct to me. I think this is the data model we agreed on

[
  {
    "versionId": "1-Qmba111111...",
    "proof": [{ ... }, { ... }]
  },
  {
    "versionId": "2-Qzmb222222...",
    "proof": [{ ... }, { ... }]
  }
]