[curves] introduce wrappers for Fp and Fq

[mimoo]

This introduces wrapper types for ark_ff::BigInteger256 as well as for our own Fp and Fq. These are supposed to add no real overhead as most functions are inlined:

Proof creation/proof creation (SRS size 2^15)                                                                          
                        time:   [3.3081 s 3.3659 s 3.4263 s]
                        change: [-0.7666% +2.6613% +5.8995%] (p = 0.15 > 0.05)
                        No change in performance detected.

(compare this with #750)

The benefit is that we can implement ocaml traits on these.

I also implemented serde serialize/deserialize on them, but I didn't manage to remove the serde_as/serde_with stuff elsewhere. I think it's doable though.

[mrmr1993]

Really nice!
As a sanity check, it would be good to make a Mina PR and make sure that CI is happy there

[mimoo]

playing with it here: MinaProtocol/mina#11906

it seems to compile the stubs. Not sure if mina compiles yet, and haven't touched the wasm part.

[mrmr1993]

See comment. Can you rework this so that it's only moving the OCaml types into this repo, and leave the rest on the Mina side?

[github-actions]

Stale pull request message

[mimoo]

probably we should try this approach first: arkworks-rs/algebra#506

EDIT: actually nevermind, this won't help us with the ocaml traits

[mimoo]

ping @mrmr1993

[mimoo]

urg actually I don't think this will solve all of our problems for serde at least... Because things like G::Scalar are not bound by Serialize/Deserialize. We might have to do the same for AffineCurve