[JS] [Benchmarking] JavaScript microbenchmarks display regressions

### Describe the bug, including details regarding any error messages, version, and platform.

I opened https://github.com/apache/arrow/issues/37657 a few months ago after noticing that the JavaScript microbenchmarks stopped working. Since then, @trxcllnt opened a PR https://github.com/apache/arrow/pull/37668 to fix them (which @assignUser merged in https://github.com/apache/arrow/pull/38500).

Now that they're running again, I noticed performance regressions in almost all of them. [See the data in Conbench.](https://conbench.ursa.dev/compare/runs/ca4cacd2979940a7a3267c91400111c3...c87efbcdc96d474e802339f05c36367c/) Many of them are over 4x slower.

Is this expected due to a known change in the benchmarking setup? If so, I can annotate these performance jumps in Conbench so that future regression alerts are more effective. If not, someone more familiar with the JS codebase should probably investigate.

### Component(s)

Benchmarking, JavaScript