False goal doesn't produce null

theory A
  goal t : false
end

produces the following proof.json:

{
  "profile": ...
  "proofs": {
    "A": {},
  }
}

instead of

{
  "profile": ...
  "proofs": {
    "A": { "t": null },
  }
}

This would allow external tools to tell whether a proof is complete by simply reading proof.json.