@profjsb so for some datasets, 10x blowup for CSV…

@profjsb so for some datasets, 10x blowup for CSV is matched by 10x speedup for reading w paratext. Some datasets no. Why, do you think?