* Add prometheus metrics to nginx unit * Run multiprocess prom collector in unit * Refactor unit metrics to use prom client properly * Ruff