Use aiofiles' AsyncBufferedReader to stream file data instead of loading each chunk into memory. Add DEFAULT_CHUNK_SIZE and typing Union/AsyncBufferedReader imports. _chunked_upload now reads and uploads slices directly from the file, passing chunk_size to _upload_chunk. _upload_chunk accepts either BytesIO or AsyncBufferedReader and uses chunk_size to set Content-length for streamed reads. upload_bytes now supports AsyncBufferedReader and enforces chunk_size limits while updating the hash function. These changes reduce memory usage and ensure correct chunk sizing for multipart uploads.