bench: add serial MSM benchmark

[mratsim]

This adds a benchmark for multi-scalar-multiplication on G1. It is the key operation for KZG commitments.

When comparing improvement over blst, using Nim is more convenient than Rust as the Rust wrapper always parallelize MSM.

The code is built in CI to make sure it doesn't go stale.

On a Ryzen 9 9950X, overclocked at 5.9GHz single core turbo.
nim c -r --cc:clang -d:danger --passC:-march=native --hints:off --warnings:off --outdir:build benchmarks/bls12381_msm_g1.nim

Constantine
CC=clang nimble bench_ec_msm_bls12_381_g1

[etan-status]

Thanks for the added benchmark!

Yeah, the double indirection in the API got me as well, one can pass multiple lists without copying all the elements into a merged single list. Also what I ended up using in #178 blst_min_pubkey_sig_core.nim.

    let  # NULL-terminated lists of pointers to lists of values
      publicKeysRef = [toCC(publicKeys[0], cblst_p1_affine), nil]
      signaturesRef = [toCC(signatures[0], cblst_p2_affine), nil]
      scalarsRef = [toCC(blindingScalars[0], byte), nil]