Refactor provable core

[mitschabaude]

The general goal here is to make the structure of core provable code more logical.

closes #1504

• get rid of several import cycles surrounding Field, by introducing a low-level stub module field-constructor.ts
• move FieldVar into its own lower level module
• define the one and only internal version of exists()
• Fix vulnerability: missing boolean check in Field.isEven()
• move Provable<T> to its own file
  • Simplify doccomments on Provable<T> 48b897c
  • Make Provable<T> a type instead of interace, because we don't use interfaces in general and they behave slightly different than types in subtle ways 82fe43b
  • also make other interfaces types, for consistency
• move some provable types into provable-types/
• move some provable types logic from bindings to provable-types/
• move some logic mostly used by mina-signer into mina-signer
• break up circuit-value.ts into separate files for CircuitValue, Struct and Unconstrained

bindings: o1-labs/o1js-bindings#258

[mitschabaude]

this was moved here from bindings/lib/provable-bigint

[mitschabaude]

this was moved here from bindings/lib/provable-snarky

[mitschabaude]

this code is the CircuitValue part of what used to be src/lib/circuit-value.ts

[mitschabaude]

this whole code used to be in field.ts

[mitschabaude]

these interfaces used to be in snarky.d.ts

[mitschabaude]

this code is the Unconstrained part of what used to be src/lib/circuit-value.ts

[ymekuria]

Grouping things into the provable-types folder and creating the unconstrained file makes things easier to understand.

[mitschabaude]

apart from existsAsync(), this now also has the exists() function we use internally a lot and that Provable.witness uses

[ymekuria]

Just curious why you decided to move this to the bindings.

[mitschabaude]

Good question! The Field module from field-bigint.ts is a wrapper around the Fp module which was always in bindings/crypto. field-bigint.ts adds other stuff to it like toBytes and toJSON methods which have nothing to do with the core finite field arithmetic. It's basically the version of Field used by mina-signer.

It's hard to track dependencies when logic is wrapped and reexported multiple times. I wanted to make it more explicit where this Fp (bigint field arithmetic) logic actually comes from, and remove the dependency on the mina-signer middle-man code where not necessary.

[ymekuria]

Ok great. That logic makes a lot of sense.

[ymekuria]

Grouping things into the provable-types folder and creating the unconstrained file makes things easier to understand.