Dual attention-based sequential auto-encoder for Covid-19 outbreak forecasting: A case study in Vietnam