FaStfact: Faster, Stronger Long-Form Factuality Evaluations in LLMs