CRAN release 2.0.6

This release changes the way the lambda tuning parameter is stored. Previously there was one lambda vector for each penalty, causing problems for elastic net-type penalties because the lambda range was off for them. The current version uses a list of vectors for lambda, one vector for each penalty, allowing for more flexibility of tuning each penalty independently