across session coding scores for each CV fold

[alexpiet]

• if a feature isn't present in one of the folds, it generates a NaN. Previously I was using nanmean() to ignore those when averaging across folds. Now I just throw a NaN, which causes some cells to crash. I could set the coding to 0 for these cells to get more data
• Separately, my code didn't have a try/except block so if an early fold had a NaN, then it didn't compute the rest of the folds.