Tests: Add test to guarantee CLI is reachable

There are tests for the CLI but these use click's `CliRunner` which
calls the commands directly from the Python API instead of calling the
actual CLI script. This would not catch if the script is not properly
installed or reachable. Here a test is added that tries to call the CLI
through a subprocess, which simulates the real situation of a user
invoking it.

This test would have caught the missing imports that were accidentally
removed by the `ruff` linter.

src/aiida_common_workflows/cli/__init__.py

@@ -1 +1,4 @@
 """Module for the command line interface."""
+from .launch import cmd_launch  # noqa: F401
+from .plot import cmd_plot  # noqa: F401
+from .root import cmd_root  # noqa: F401

tests/cli/test_root.py

@@ -0,0 +1,45 @@
+"""Tests for CLI commands."""
+from __future__ import annotations
+
+import subprocess
+
+import click
+import pytest
+from aiida_common_workflows.cli import cmd_root
+
+
+def recurse_commands(command: click.Command, parents: list[str] | None = None):
+    """Recursively return all subcommands that are part of ``command``.
+
+    :param command: The click command to start with.
+    :param parents: A list of strings that represent the parent commands leading up to the current command.
+    :returns: A list of strings denoting the full path to the current command.
+    """
+    if isinstance(command, click.Group):
+        for command_name in command.commands:
+            subcommand = command.get_command(None, command_name)
+            if parents is not None:
+                subparents = [*parents, command.name]
+            else:
+                subparents = [command.name]
+            yield from recurse_commands(subcommand, subparents)
+
+    if parents is not None:
+        yield [*parents, command.name]
+    else:
+        yield [command.name]
+
+
+@pytest.mark.parametrize('command', recurse_commands(cmd_root))
+@pytest.mark.parametrize('help_option', ('--help', '-h'))
+def test_commands_help_option(command, help_option):
+    """Test the help options for all subcommands of the CLI.
+
+    The usage of ``subprocess.run`` is on purpose because using :meth:`click.Context.invoke`, which is used by the
+    ``run_cli_command`` fixture that should usually be used in testing CLI commands, does not behave exactly the same
+    compared to a direct invocation on the command line. The invocation through ``invoke`` does not go through all the
+    parent commands and so might not get all the necessary initializations.
+    """
+    result = subprocess.run([*command, help_option], check=False, capture_output=True, text=True)
+    assert result.returncode == 0, result.stderr
+    assert 'Usage:' in result.stdout