Makes all CaretList's regression predictions be numeric

[Malhadas]

Caret not always returns numeric or factor predictions, instead sometimes it returns them as character wich results in errors in the caretEnsemble and caretStack functions. The underlying packages that implement the models themselves are in constant change and cannot be 100% trusted to always give the numeric values caretEnsemble and caretStack expect. In order to solve this issue, the line if(!is.null(model)) model$pred$pred <- as.numeric(model$pred$pred) can be added on the caretList function.

[lintr-bot]

R/caretList.R:236:71: style: Trailing whitespace is superfluous.

​    if(!is.null(model)) model$pred$pred <- as.numeric(model$pred$pred)               
                                                                      ^~~~~~~~~~~~~~~

R/caretList.R:246:1: style: Trailing whitespace is superfluous.

​      
^~~~~~

R/caretList.R:247:1: style: Trailing whitespace is superfluous.

​      
^~~~~~

[coveralls]

Coverage decreased (-31.6%) to 65.139% when pulling e80fad56e62ecb855aa63eb4e7e8981f2efb604e on Malhadas:patch-1 into 48e6ecd65fa53e3255162a867454fcd289b98307 on zachmayer:master.

[ran88dom99]

I have this problem. https://github.com/zachmayer/caretEnsemble/issues/210

[Malhadas]

I have this problem. #210

The solution I proposed worked for me and has not given me any problem yet