chore: change rust kzg nif to direct elixir bindings

[BlazeWasHere]

Motivation

Removes the overhead from elixir -> rust (via rustler) -> c and now does elixir -> c via these elixir bindings.

Description

Removes usage of the c-kzg Rustler NIF and replaces it with direct c-kzg bindings instead.

Getting speedups of ~40x with verify_blob_kzg_proof with this change

Note: This NIF loads the trusted setup (settings) once, so essentially closes #848

# Before
~> git fetch origin 4111b5d
~> mix run bench/kzg.exs
Operating System: macOS
CPU Information: Apple M2
Number of Available Cores: 8
Available memory: 16 GB
Elixir 1.18.4
Erlang 28.0
JIT enabled: true

Benchmark suite executing with the following configuration:
warmup: 2 s
time: 5 s
memory time: 0 ns
reduction time: 0 ns
parallel: 1
inputs: none specified
Estimated total run time: 35 s

Benchmarking blob_to_kzg_commitment ...
Benchmarking compute_blob_kzg_proof ...
Benchmarking compute_kzg_proof (z) ...
Benchmarking verify_blob_kzg_proof ...
Benchmarking verify_blob_kzg_proof_batch (4) ...
Calculating statistics...
Formatting results...

Name                                      ips        average  deviation         median         99th %
verify_blob_kzg_proof                   16.29       61.38 ms     ±0.89%       61.41 ms       63.65 ms
verify_blob_kzg_proof_batch (4)         15.22       65.69 ms     ±9.04%       64.71 ms      107.64 ms
compute_kzg_proof (z)                    9.86      101.42 ms     ±7.65%       99.41 ms      145.88 ms
compute_blob_kzg_proof                   9.10      109.92 ms     ±9.62%      106.37 ms      165.99 ms
blob_to_kzg_commitment                   8.96      111.61 ms     ±5.78%      111.75 ms      141.87 ms

Comparison: 
verify_blob_kzg_proof                   16.29
verify_blob_kzg_proof_batch (4)         15.22 - 1.07x slower +4.31 ms
compute_kzg_proof (z)                    9.86 - 1.65x slower +40.04 ms
compute_blob_kzg_proof                   9.10 - 1.79x slower +48.54 ms
blob_to_kzg_commitment                   8.96 - 1.82x slower +50.22 ms

# Now apply changes
~> git fetch origin 815a2c
~> mix deps.get
~> mix run bench/kzg.exs --mode db
[...]
Name                                      ips        average  deviation         median         99th %
verify_blob_kzg_proof                  634.65        1.58 ms    ±12.26%        1.56 ms        1.72 ms
verify_blob_kzg_proof_batch (4)        216.31        4.62 ms     ±1.71%        4.60 ms        5.02 ms
blob_to_kzg_commitment                  25.59       39.08 ms     ±5.46%       38.47 ms       55.22 ms
compute_kzg_proof (z)                   25.54       39.16 ms     ±2.14%       39.03 ms       45.09 ms
compute_blob_kzg_proof                  25.37       39.41 ms     ±2.47%       39.20 ms       44.52 ms

Comparison: 
verify_blob_kzg_proof                  634.65
verify_blob_kzg_proof_batch (4)        216.31 - 2.93x slower +3.05 ms
blob_to_kzg_commitment                  25.59 - 24.80x slower +37.50 ms
compute_kzg_proof (z)                   25.54 - 24.85x slower +37.58 ms
compute_blob_kzg_proof                  25.37 - 25.01x slower +37.84 ms

[BlazeWasHere]

not sure if there should be another flag to skip compiling the nif?

[BlazeWasHere]

I am pretty sure proof size is 384 and not 768, but feel free to revert

[BlazeWasHere]

not sure i should add precompiling the nif here instead?