More store object info json cleanup

[Ericson2314]

Motivation

See each commit for details.

Context

JSON cleanup #13570

---

Add 👍 to pull requests you find important.

The Nix maintainer team uses a GitHub project board to schedule and track reviews.

[edolstra]

FWIW, Determinate Nix has adopted the following format for serializing hashes in a structured way:

{"algo": "sha256", "base16": "abcdef..."}

https://github.com/DeterminateSystems/nix-src/blob/d2142024ca3688e0c62c5b7d567eca14b482318c/src/libutil/hash.cc#L495-L501

It makes implementations simpler if we have only one hash encoding format (i.e. base-16) in JSON.

[Ericson2314]

We can perhaps change our format. (Though do note that that was in a different PR much before this one, this one is just using the pre-existing format.)

It may help to know that my intent was not to support multiple hash formats --- mine always outputs base-64 unconditionally --- but make sure that we only support one hash format at a time.

For example, yours will parse

{
  "algo": "sha256",
  "base16": "abcdef...",
}

and might someday parse

{
  "algo": "sha256",
  "base64": "FePFYIlMuycIXPZbWi7LGEiMmZSX9FMbaQenWBzm1Sc="
}

but if given the union of things it parses at that point:

{
  "algo": "sha256",
  "base16": "abcdef...",
  "base64": "FePFYIlMuycIXPZbWi7LGEiMmZSX9FMbaQenWBzm1Sc="
}

but what does it do in that situation? Parse both and make sure they match?

mine will parse both

{
  "algorithm": "sha256",
  "format": "base64",
  "hash": "FePFYIlMuycIXPZbWi7LGEiMmZSX9FMbaQenWBzm1Sc="
}

and

{
  "algorithm": "sha256",
  "format": "base16",
  "hash": "abcdef..."
}

(though I wouldn't mind doing some anti-"Postel's law" and just refusing to parse format != baseWhateverWeChoose.)

but even if it does parse both, there is no union: the schema makes very clear that one can only pass one hash at a time.

[edolstra]

The field name base16 is not intended to denote that there will ever be another supported encoding. It could also be named hash. But calling it base16 makes it self-documenting.

[Ericson2314]

(Note that the hash instance was added in #14307)

[Ericson2314]

Per #14532 I think this PR is unblocked, because this PR is about using the hash format for JSON, not making it. After the issue is fixed (before next release) all this JSON will be regenerated accordingly, but not of the C++ this PR touches will need to be touched.