Scalability comparison between HISAT-3N, Bismark, BS-Seeker2, and BSMAP on 10 million simulated 100-bp paired-end BS-seq reads (0.2% per-base sequencing error rate)