Package up CSV.detect

[kescobo]

From discussion on slack, started with me:

Is there a way to tell CSV.read() not to try to unify types for a column? I've got some long form data where the values are a mix of Ints, Floats, and strings. Right now it's just parsing everything as a string and I have to do a second pass to correct everything

@quinnj said:

Ok, so one thing we could do that I've been thinking of doing anyway (since the introduction of the CSV.Rows type) is to package up our CSV.detect machinery so you could just do CSV.detect(str::String), and it would use all the same logic CSV.jl already uses when parsing a file to detect column types. Then you could use CSV.Rows and just call detect to parse the types, or in your case, use CSV.File and apply detect afterwards on the string column. Would you mind opening an issue for this? I don't think it should be too hard to do.