refactor(bench): use node:util parseArgs for CLI parsing

[TylerVigario]

Summary

Replace the hand-rolled flag-loop in bench/index.js with parseArgs from node:util — available since Node 22, which is engines.node's floor.

The previous parser was ~62 lines of imperative "skip the next arg if the previous was a flag" bookkeeping. The new version is a flat options declaration:

parseArgs({
  options: {
    lang: { type: 'string', multiple: true },
    language: { type: 'string', multiple: true },
    function: { type: 'string', multiple: true },
    fn: { type: 'string', multiple: true },
    value: { type: 'string' },
    save: { type: 'boolean' },
    // ...
  },
  allowPositionals: true,
  strict: true
})

Net +48 / -62. Each flag's name, type, and multi-occurrence behavior is visible at a glance.

Behavior change

Unknown flags now throw via parseArgs's strict: true mode. Previously a typo like --lng en-US would silently fall through and the bench would run with default arguments. Failing fast is the right tradeoff for a benchmark CLI — you find out at parse time, not after a 30-second run that benchmarked the wrong thing.

All other behaviors preserved: --lang/--language aliases, comma-separated values within a single flag, repeating flags (--lang en --lang fr), positionals as language codes, unknown function name validation with the existing red error message.

Test plan

• node bench/index.js --help — works
• node bench/index.js en-US (positional) — runs single language
• node bench/index.js --bogus-flag — fails fast with parseArgs error
• node bench/index.js --function nonexistent — preserves friendly error + exit 1
• npx standard bench/index.js — lint clean

🤖 Generated with Claude Code