AMuRD: Annotated Multilingual Receipts Dataset for Cross-lingual Key Information Extraction and Classification