Read multiple rows at a time to improve read performance

[keichi]

This improves read IO performance by more than 10x:

• Subject6 (2.8GB):
  • before: Read input dataset (8528 rows, 101729 columns) in 88884.1 [ms]
  • after: Read input dataset (8528 rows, 101729 columns) in 8462.89 [ms]
• Subject11 (7.6GB):
  • before: Read input dataset (3780 rows, 92538 columns) in 34381.2 [ms]
  • after: Read input dataset (3780 rows, 92538 columns) in 3018.65 [ms]