R Faster way to lag by entire group using data.table

Question

I have a simple data.table as follows-

  ID = c(rep("A", 1000), rep("B", 1000), rep("C", 1000), rep("D", 1000))
  val = re