refactor: Compile(ecc.ID) -> Compile(field *big.Int)

[gbotrel]

This PR makes frontend.Compile hints witness and cs packages dependent on a field instead of being dependent on a curveID.

This PR adds a tinyfield package (modulus == 47), for test purposes.

I believe the only algorithmic change is witness Serialization; instead of using curve.Encoder it directly uses fr.Element [Set]Bytes() methods.

• build: update to latest gnark-crypto
• refactor: frontend.Compile(ecc.ID) -> frontend.Compile(field.Field)
• refactor: field.Field -> *big.Int
• build: update to latest gnark-crypto
• feat: added internal/tinyfield
• refactor: NewWitness(ecc.ID) -> NewWitness(field big.Int)
• feat: build R1CS with tinyfield
• test: ignore range circuit for tinyfield r1cs serialization
• refactor: hint now takes *big.Int scalar field instead of curveID

[ivokub]

I think that in general the changes make sense. I'm just thinking that maybe it would be better to name the modulus variable as primeFieldModulus etc? If it is field, then may be ambiguous?

[drakstik]

Hey, I know this is a closed issue, but it was not updated in your Go package documentations https://pkg.go.dev/github.com/consensys/gnark#section-readme. Specifically, the first example given of a cubic equation. Would save time for beginners to just add the ecc.BN254.ScalarField() to the example! I had to look around the GitHub for this, which is good practice, but I'd rather save other's time just a little.

Thanks for all the hard work!