Cache hash outputs

We can likely optimize the computation of row/col roots by up to 2x if we cache the outputs of leafs, because right now we are hashing all the data chunks twice, once to compute the row roots, and once to compute the col roots.

Start by locating the row-root and column-root computation and trace where leaf data chunks are hashed. Compare the current work in both paths, then verify that leaf hash outputs are reused without changing the resulting roots; add or run relevant tests and benchmarks to confirm correctness and the expected reduction in hashing work.

Written by the indexing model from the issue text.