Inefficient executeHashComputations of hashtree

benchmark does not show great `executeHashComputations()` performance on Linux, for example in https://github.com/ChainSafe/ssz/actions/runs/14090676972/job/39466309168

```
hashtree
✔ getHashComputations 369.1343 ops/s 2.709041 ms/op x1.226 119 runs 12.2 s
✔ executeHashComputations 100.3185 ops/s 9.968249 ms/op x0.[90](https://github.com/ChainSafe/ssz/actions/runs/14090676972/job/39466309168#step:6:91)5 50 runs 4.91 s
✔ get root 65.38101 ops/s 15.29496 ms/op x1.030 12 runs 2.27 s
```

`executeHashComputations` is only 50% faster than `get root`, on our typical Linux node it's only 30% faster

however on my Macbook, it's 150% faster.

Start by running the hashtree benchmark on Linux and compare executeHashComputations() with getHashComputations() and get root, using the linked GitHub Actions run as a baseline. Trace the executeHashComputations() path to identify the source of the performance gap, then confirm the benchmark shows a meaningful improvement on Linux.

Written by the indexing model from the issue text.