Add L1 reg support(OWLQN) for VLogisticRegression

add elasticNet param for VLogisticRegression
adjust OWLQN constructor for reg param (withIntercept support)
add mapPartitionsWithIndex and compressed method for DistributedVector
add V-LOR model compression after training complete.(because OWLQN will generate sparse model and it is easy to compress.)
add testcases for V-LOR elasticNet param.

yanboliang / spark-vlbfgs