Gravar-mail: Comparing Multimodal LLMS and Traditional Neural Networks for Table Extraction From PDFs and Images: An Evaluation of Structure and Content Extraction from Table in Images