Add CLI smoke tests for all subcommands

[dreamlessx]

Description

Add smoke tests that verify all CLI subcommands parse arguments correctly and produce expected output without requiring GPU or model weights.

Target subcommands

• python -m landmarkdiff --help — verify exit code 0 and usage text
• python -m landmarkdiff predict --help — verify procedure/intensity flags shown
• python -m landmarkdiff serve --help — verify host/port flags shown
• python -m landmarkdiff evaluate --help — verify metric flags shown
• python -m landmarkdiff export --help — verify format flags shown

Example test

import subprocess

def test_help_returns_zero():
    result = subprocess.run(
        ["python", "-m", "landmarkdiff", "--help"],
        capture_output=True, text=True,
    )
    assert result.returncode == 0
    assert "usage" in result.stdout.lower()

Motivation

CLI entry points at 61% coverage. Smoke tests are the cheapest way to cover argparse setup code and catch import errors.

Acceptance criteria

• tests/test_cli_smoke.py created with tests for all subcommands
• Tests run without GPU/models (help flags only)
• Tests pass on Python 3.10, 3.11, 3.12