Add option to report classification performance for each fold in crossval()

[stefan-mueller]

Could we add an option for crossval() to report the results from each fold? This could be useful if users want to get the mean, minimum and maximum scores across the folds.

Example: if by_class = FALSE, the resulting data frame would have five observations for a five-fold cross validation; if by_class = TRUE this number of observations is multiplied by the number of classes.

Not sure how to call the argument – maybe summarize_folds = TRUE/FALSE?