sharkdp/hyperfine ? reverse-engineered prompt

Build me a Rust terminal app that lets me benchmark shell commands and compare how long they take. I want to run one command or several commands, have the tool repeat them enough times to get reliable numbers, then show a clean summary with mean time, min, max, standard deviation, and relative speed.

It should support warmup runs, running a setup or preparation command before timing, and optional cleanup after. It should keep showing progress while the benchmark is running, notice obvious outliers, and try to avoid counting shell startup time when possible. I also want to test variations of a command by changing a parameter, like number of threads or compiler choice.

Please include export options for CSV, JSON, Markdown, and AsciiDoc so results can be shared or plotted later. Make it work across common platforms, include helpful terminal output and help text, and add tests for the important behavior.