Add microbenchmarks for computing a chunk group hash

[rklaehn]

It looks like computing a chunk group hash is faster in abao than in bao-tree. To fix this we should have a microbenchmark comparing the two, and then obviously fix it.

See

n0-computer/iroh#1288

[rklaehn]

It looks like this was a false hope. The binary where the outboard computation is super fast is already using bao-tree, at least from the symbol table...

abao/bao_bin on  master [$!?] is 📦 v0.12.1 via 🦀 v1.71.0 
❯ nm /Users/rklaehn/bin/iroh | grep bao_tree | wc -l
     221

abao/bao_bin on  master [$!?] is 📦 v0.12.1 via 🦀 v1.71.0 
❯ nm /Users/rklaehn/bin/iroh | grep abao | wc -l    
       0

[rklaehn]

The issue of the fastest way to compute a chunk group hash is solved in BLAKE3-team/BLAKE3#329 / the iroh-blake3 crate and whatever @oconnor663 will come up with to solve this long term.