Human vs. generative artificial intelligence in writing assessment: investigating feedback alignment, score validity, and teacher agency