chore: convert benchmark for single cell verification to batch verification

[jtraglia]

Noticed this wasn't being benchmarked & we're about to remove single cell verification.

Here are the results from my system. I was a little surprised that num_threads didn't make a difference. Is that normal?

verify_cell_kzg_proof_batch - NUM_THREADS: 1
                        time:   [11.931 ms 11.954 ms 11.977 ms]
verify_cell_kzg_proof_batch - NUM_THREADS: 4
                        time:   [11.915 ms 11.951 ms 11.999 ms]
verify_cell_kzg_proof_batch - NUM_THREADS: 8
                        time:   [11.940 ms 11.965 ms 11.992 ms]
verify_cell_kzg_proof_batch - NUM_THREADS: 16
                        time:   [11.931 ms 11.958 ms 11.987 ms]
verify_cell_kzg_proof_batch - NUM_THREADS: 32
                        time:   [11.897 ms 11.920 ms 11.944 ms]

[kevaundray]

I was a little surprised that num_threads didn't make a difference.

The benchmarks seem to only be using a single commitment, which saves quite a bit of MSMs -- currently the FFT algorithm is not parallelized, so I think this is expected, I would expect more of a discrepancy if we were to use different commitments

[kevaundray]

The choosenim github action has been flaky over the last couple of months -- going to merge since it passes for nim on ubuntu and windows, and its only modifying benchmark code