Introduce halo2_proofs::circuit::Value

[str4d]

This replaces every usage of Option<V> for representing witnessed values that may not have a known value (e.g. advice cells during keygen).

This is a very big breaking change to the public API, but does not alter circuit compatibility.

Closes #509.

[str4d]

Force-pushed to fill in the changelogs. The breaking changes to the public API aren't quite as bad as I thought! Still large, but enumerable.

[str4d]

zcash/orchard#336 has the migration of orchard to the new APIs.

[CPerezz]

Looks really good!! Just have some questions regarding the ops traits impls for Value.

Feels much better now avoiding a lot of unwraps and ok_or(Error::Synthesis). The readability will be improved quite a lot also IMO.
Thanks a lot for working on this @str4d 🥳

[daira]

utACK

Consider shortening uses of Value::known by importing it and using known(...).

[str4d]

Consider shortening uses of Value::known by importing it and using known(...).

This doesn't actually work; Rust lets you import enum cases, but not struct methods. We'd need a standalone method instead, which would then be weird as it would lack context (halo2_proofs::circuit::known).

[therealyingtong]

utACK