Minimal memory allocation for ViewDU batch hash

[twoeths]

Is your feature request related to a problem? Please describe.

batch hash work is in progress, this is how it looks in a lodestar node every time we get through epoch transition

• 6h chart

• 10m chart

looks like memory allocation is an issue there and it caused some spiked hash_tree_root times

Describe the solution you'd like

right now we allocate memory every time we compute ViewDU.hashTreeRoot(), find a way to do minimal memory allocation and reuse it over time

Additional context

• tested branch: te/batch_hash_tree_root