Training f5c/nanopolish models for nanopore R10.4.1 chemistry

We trained models for R10.4.1 for use with f5c eventalign and call-methylation. Given f5c relies on nanopolish models, we use nanopolish for training such models. We sequenced the human methylated and non-methylated (WGA) DNA datasets from zymo dna methylation standards (D5013). One PromethION flowcell was used for each dataset and we obtained ~7X depth for the non-methylated and ~9X for the methylated. The whole genome amplified non-methylated data with no base modifications is used for training neucleotide model. Both methylated (only CG methylated in this Zymo dataset) and non-methylated are needed for training a methylation model in CpG context. More information is detailed at https://hasindu2008.github.io/f5c/docs/r10train.