Inefficient executeHashComputations of hashtree

[twoeths]

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.

[twoeths]

benchmark on #480 shows how well hashtree is but it does not reflex in executeHashComputations()

• our regular Linux node

 hashtree
      ✔ hash 2 32 bytes Uint8Array 500000 times - hashtree                  5.833505 ops/s    171.4235 ms/op        -         12 runs   22.3 s
      ✔ batch hash 16 x 64 Uint8Array 31250 times - hashtree                38.42098 ops/s    26.02745 ms/op        -         12 runs   3.64 s

 hashtree
    ✔ getHashComputations                                                 675.1451 ops/s    1.481163 ms/op   x0.997         65 runs   11.7 s
    ✔ executeHashComputations                                             111.0028 ops/s    9.008785 ms/op   x0.869         60 runs   4.03 s
    ✔ get root                                                            77.11619 ops/s    12.96745 ms/op   x0.968         16 runs   2.13 s

• my Macbook

  hashtree
      ✔ hash 2 32 bytes Uint8Array 500000 times - hashtree                  9.301045 ops/s    107.5148 ms/op   x0.886         12 runs   14.0 s
      ✔ batch hash 16 x 64 Uint8Array 31250 times - hashtree                51.79010 ops/s    19.30871 ms/op   x0.987         12 runs   2.90 s

  hashtree
    ✔ getHashComputations                                                 2303.723 ops/s    434.0800 us/op   x0.979        168 runs   13.1 s
    ✔ executeHashComputations                                             263.6497 ops/s    3.792911 ms/op   x0.999        128 runs   3.45 s
    ✔ get root                                                            112.4302 ops/s    8.894407 ms/op   x1.008         31 runs   1.61 s

platform	hashtree batch capability	executeHashComputations
typical Linux	6.58x	1.44x
Macbook	5.57x	2.35x

[twoeths]

the overhead in executionHashComputations mainly comes from HashObject to/from Uint8Array conversion
this can only be avoided if we model 32 bytes as HashObject as before which can only be achieved with Bun ChainSafe/lodestar#5857