Suppress output in CLI tests

When testing code paths that write to stderr/stdout or call process.exit (e.g. error handling, help, version output), avoid manually hijacking process.exit, console.error, or spying on process.stderr.write. These approaches are brittle, verbose, and pollute test output. Instead, use the library’s built-in configuration hooks to suppress output and convert exits into catchable errors, and prefer shared test helpers/assertion utilities over ad-hoc mocks.

For example, instead of:

var oldProcessExit = process.exit;
var oldConsoleError = console.error;
process.exit = function() { exceptionOccurred = true; throw new Error(); };
console.error = function() {};

use:

const program = new commander.Command();
program
  .exitOverride()
  .configureOutput({ writeErr: () => {} });

assert.throws(
  () => program.parse(['node', 'test', '-m']),
  (err) => {
    assert.ok(err instanceof commander.CommanderError);
    assert.equal(err.exitCode, 1);
    assert.equal(err.code, 'commander.unknownOption');
    return true;
  },
);

Where many tests share this setup, factor it into a shared helper (e.g. createTestCommand()) so tests stay concise and consistent, and output-suppression logic isn’t duplicated or forgotten. This keeps test output clean, makes assertions explicit and reliable, and avoids leaking global state changes (like reassigned process.exit) between tests.