Optimize PeerDAS

[ArtiomTr]

In this PR, several optimizations are ported from rust-eth-kzg:

1. Batch multiplication precomputation - doing precomputations after loading trusted setup
2. Using HashMap for deduplicate_commitments
3. Using HashSet to find missing cell indices in cell recovery
4. Batch inverse

Parallel benchmark times, with avg. CPU load:

	rust-kzg	avg load
compute_cells_and_kzg_proofs	48.513 ms	27%-30%
recover_cells_and_kzg_proofs (50% missing)	59.624 ms	25-30%
recover_cells_and_kzg_proofs (25% missing)	61.589 ms	25-30%
recover_cells_and_kzg_proofs (12.5% missing)	57.518 ms	25-30%
recover_cells_and_kzg_proofs (1 missing)	55.628 ms	25-30%
recover_cells_and_kzg_proofs (2 missing)	56.000 ms	25-30%
recover_cells_and_kzg_proofs (3 missing)	55.269 ms	25-30%
recover_cells_and_kzg_proofs (4 missing)	54.523 ms	25-30%
recover_cells_and_kzg_proofs (5 missing)	57.823 ms	25-30%
verify_cell_kzg_proof_batch	148.35 ms	37-40%
verify_cell_kzg_proof_batch (rows)/1	9.1205 ms	18-20%
verify_cell_kzg_proof_batch (rows)/2	12.693 ms	20-22%
verify_cell_kzg_proof_batch (rows)/4	16.208 ms	23-24%
verify_cell_kzg_proof_batch (rows)/8	24.932 ms	27-30%
verify_cell_kzg_proof_batch (rows)/16	41.965 ms	30-33%
verify_cell_kzg_proof_batch (rows)/32	76.117 ms	30-33%
verify_cell_kzg_proof_batch (rows)/64	147.48 ms	34%
verify_cell_kzg_proof_batch (columns)/1	9.0926 ms	20-22%
verify_cell_kzg_proof_batch (columns)/2	9.8894 ms	22-24%
verify_cell_kzg_proof_batch (columns)/4	12.491 ms	26-27%
verify_cell_kzg_proof_batch (columns)/8	16.998 ms	27-28%
verify_cell_kzg_proof_batch (columns)/16	27.090 ms	28-30%
verify_cell_kzg_proof_batch (columns)/32	42.328 ms	29-30%
verify_cell_kzg_proof_batch (columns)/64	75.353 ms	32-34%
verify_cell_kzg_proof_batch (columns)/128	153.58 ms	38-40%

[sauliusgrigaitis]

Is this still in draft mode?

[ArtiomTr]

Yeah, I've replaced std HashMap & HashSet with hashbrown because of no_std support, so I want to re-run benchmarks to make sure that performance haven't degraded