Create a link to DotsOCR

Use direct import of modeling_dots_ocr_vllm in vLLM entrypoint

Seperate everything into files