detect_dm_csv return type "logical" for empty columns

[djvanderlaan]

When a CSV file has an empty column (or when the column is empty for the rows used), the column type returned is logical which is not a valid column type.

The following code reproduces the problem
lines <- c("1;;A", "2;;A", "3;;B", "4;;B")
writeLines(lines, "test.csv")
dm <- detect_dm_csv("test.csv", sep=";")
dm

For unsupported column types we probably need to generate a warning and use type "string". Type "string" will always work and read in the data. Perhaps we also need to detect empty columns although it is not clear what we should do with those as it is impossible to determine the correct type.