These are not intended for use by users. lambda.interp does linear
interpolation of the lambdas to obtain a prediction at a new point s.
glmnet_softmax does the classification for multinomial models.
nonzeroCoef determines in an efficient manner which variables are
nonzero in each fit. jerr prints out error messages from the fortran.
plotCoef is called by the plot method for glmnet
objects. check_dots is used in coef and predict with
argument exact=TRUE, to make sure user supplies original data used to
fit the "glmnet" object.