sumFlat - Encodes sum types with single record fields in one object

[m-bock]

Let's say we have a PureScript Sum type, each case has either no fields or a single record field:

data Sample
  = Foo
  | Bar { errors ∷ Int }
  | Baz { active ∷ Boolean, name ∷ String, count ∷ Int }

This PR introduces two new functions : sumFlat and sumFlatWith. They works similar to sum and sumWIth.
However for each constructor a JPropCodec (or unit) must be provided.

sumFlat uses a default encoding, whereas sumFlat can be passed custom encoding options.

In the custom options a type level symbol can be given that represents a case discriminator in the JSON. (usually "type", "kind", "tag"...)

import Data.Codec.Argonaut as C
import Data.Codec.Argonaut.Record as CR
import Data.Codec.Argonaut.Sum as CS

codecSample ∷ JsonCodec Sample
codecSample = CS.sumFlatWith { tag: Proxy @"tag"} "Sample"
  { "Foo": unit
  , "Bar": CR.record { errors: C.int }
  , "Baz": CR.record { active: C.boolean, name: C.string, count: C.int }
  }

This codec can then be used as usual:

C.encode codecSample (Baz { active: true, name: "hello", count: 42 })

It will encode as follows:

{
  "tag": "Baz",
  "active": true,
  "count": 42,
  "name": "hello"
}

It's verified at compile time that the record fields don't conflict with the custom tag. So in this example, if you exchange "name" with "tag" in the type definition, the compilation will fail.

[garyb]

Awesome, this looks great. Thanks so much for working on it!