Add XLM-RoBERTa model - #136

{:ok, model_info} = Bumblebee.load_model({:hf, "papluca/xlm-roberta-base-language-detection"})
{:ok, tokenizer} = Bumblebee.load_tokenizer({:hf, "papluca/xlm-roberta-base-language-detection"})
serving = Bumblebee.Text.text_classification(model_info, tokenizer)
Nx.Serving.run(serving, "Kraków w śniegu.")
#=> %{
#=>   predictions: [
#=>     %{label: "pl", score: 0.9958330988883972},
#=>     %{label: "el", score: 4.284092574380338e-4},
#=>     %{label: "bg", score: 4.1166096343658864e-4},
#=>     %{label: "es", score: 3.488743386697024e-4},
#=>     %{label: "de", score: 3.1762008438818157e-4}
#=>   ]
#=> }