Added univariate functionality

[lynnyi]

In glmnet(x, y), allow x to be a vector instead of a matrix in order to perform univariate regression.

[emallickhossain]

I don't know if this will every get merged since this is just the CRAN mirror repo. I don't know where the active source code is stored. My guess is that Trevor has it in a private repo.

[marjoleinF]

If one wants to do univariate regression with glmnet, it might be easiest to just add a vector of zeros:

x <- rnorm(100)
y <- x + rnorm(100)
cv.glmnet(x = cbind(0, x), y = y)