lexpand: In pop.haz repeat the risk of death of the first period backwards in time if needed

[mattirantanen]

This would be consistent behaviour as lexpand replicates the latest year of pophaz data when necessary. This would avoid the NA values in variable pop.haz.

Example where pop.haz is NA with a warning:

dat <- data.table(bi_date = 1890, dg_date = 1895, ex_date = 2000, status = 1, sex = 1)
    x3 <- lexpand(dat, status = status != 0,
                  birth = bi_date, entry = dg_date, exit = ex_date,
                  pophaz=popmort)