chore: Refactors CPU MSM operations using halo2curves library

- Expanded the `msm` module within the `provider/util` directory and introduced a new function, `cpu_best_msm`.
- Changed the curves library dependency in `bn256_grumpkin.rs` from `pasta_curves` to `halo2curves`,
- Removed the `msm.rs` file along with two associated functions `cpu_msm_serial` and `cpu_best_msm` used for non-GPU accelerated operations, and all related tests.
- Reorganized the import of `CurveAffine, CurveExt` from the `halo2curves` library in `provider/mod.rs`.

Fixes argumentcomputer#193

Cargo.toml

@@ -54,7 +54,7 @@ pasta-msm = { git = "https://github.com/lurk-lab/pasta-msm", branch = "dev", ver
 grumpkin-msm = { git = "https://github.com/lurk-lab/grumpkin-msm", branch = "dev", features = ["dont-implement-sort"] }
 
 [target.'cfg(not(target_arch = "wasm32"))'.dependencies]
-halo2curves = { version = "0.5.0", features = ["bits", "derive_serde"] }
+halo2curves = { version = "0.5.0", features = ["bits", "derive_serde", "multicore"] }
 
 [target.'cfg(target_arch = "wasm32")'.dependencies]
 # see https://github.com/rust-random/rand/pull/948

src/provider/bn256_grumpkin.rs

@@ -9,10 +9,10 @@ use ff::{FromUniformBytes, PrimeField};
 use group::{cofactor::CofactorCurveAffine, Curve, Group as AnotherGroup};
 #[cfg(any(target_arch = "x86_64", target_arch = "aarch64"))]
 use grumpkin_msm::{bn256 as bn256_msm, grumpkin as grumpkin_msm};
+// Remove this when https://github.com/zcash/pasta_curves/issues/41 resolves
+use halo2curves::{CurveAffine, CurveExt};
 use num_bigint::BigInt;
 use num_traits::Num;
-// Remove this when https://github.com/zcash/pasta_curves/issues/41 resolves
-use pasta_curves::arithmetic::{CurveAffine, CurveExt};
 use rayon::prelude::*;
 use sha3::Shake256;
 use std::io::Read;

src/provider/util/mod.rs

@@ -1,3 +1,12 @@
 /// Utilities for provider module
 pub(in crate::provider) mod fb_msm;
-pub(in crate::provider) mod msm;
+pub mod msm {
+  use halo2curves::msm::best_multiexp;
+  use halo2curves::CurveAffine;
+
+  // this argument swap is useful until Rust gets named arguments
+  // and saves significant complexity in macro code
+  pub fn cpu_best_msm<C: CurveAffine>(bases: &[C], scalars: &[C::Scalar]) -> C::Curve {
+    best_multiexp(scalars, bases)
+  }
+}