Enable using kwargs for selecting pad-to-max-length strategy for tokenizer in embeddings

Enable using kwargs for selecting pad-to-max-length strategy for tokenizer in embeddings #1139