Upstream various benchmarking changes

[ilan-gold]

Right now we have a heavily modified fork of zarrs_tools. A couple of things need to be upstreamed:

• Environment management.. There is no command for how to create an env, what packages are needed etc.
• System specifics API for swapping out commands for timing and cache busting (and anything else that might be system-specific)
• Small Datasets + CI The benchmarks probably should run on CI at least in a small case. I'm had a lot of trouble getting them to work on both Rocky Linux and my ubuntu. Rocky Linux did not like the Popen (I had to pass in shell=True) and ubuntu was complaining about CLI downcasting
• Parallelization Currently, the zarrs-python is parallelized by default so benchmark is a bit wonky and not really "fair." See this issue.

[ilan-gold]

From Phil:

• Preprend time command with bash -c or something so we don't have to rely on the file itself being present (it isn't on our HPC)