fix: refactor get_dataset_file to obtain total rows from metadata

[Irozuku]

This pull request refactors how the total number of rows is determined in the get_dataset_file endpoint. Instead of incrementally counting rows while reading batches from the Arrow file, it now retrieves the total row count using the get_dataset_info function, which simplifies the code and fixes the count of rows

Dataset row count calculation:

• Removed the on-the-fly calculation of total_rows during batch iteration, eliminating the need to incrementally count rows while processing the file. [1] [2]
• Added a call to get_dataset_info to retrieve the total number of rows after batch processing, streamlining the logic and making the code easier to maintain.

Before

After