Flexible col_types specification with tidyselect

[jxu]

Feature request, for example in read_csv to have all columns that start with "abc" be interpreted as col_factor(). Right now I do it after import with mutate/across. I'm not sure what the syntax would be like.