duplicate text_config and text_config_dict and for vision_config

add onnx model

fix DFN token_embedding size mismatch

add preprocesser_config

from Apple/DFN-public

initial commit