nested CV fails if X_extra is not included in splitter evaluate call

[bafflerbach]

A student in skunkworks had this error when running nested CV while having not set X_extra during data loading leading to X_extra being empty. This can be reproduced by modifying the final tutorial splitter to not include the X_extra input in the splitter call itself, or probably by just not setting X_extra during data loading (and subsequently dropping the meta data from feature set manually which i think is what the student did).

The error is:
"Nonetype" object has no attribute columns
from line 603 in data_splitters.py

The error occurs because the code is assuming X_extra exists and has columns when trying to write output files. So there should either be a check earlier to make sure writing this file is appropriate or a warning to make sure to include it, maybe both?

[bafflerbach]

An additional thing the student noted is that the error seemed sensitive to whether or not the plots input had additional types included. I'm not able to reproduce this however so may just be an error in their script and not a bug in and of itself.

plots = ["Error"]
did not result in the same error

plots = ["Error","Histogram","Scatter"]
Still had the error even when X_extra was set