Changes to `shift` cause backwards naming issues

[daynefiler]

This is totally related to #3223 and the new shift behavior from #1708. I just want to point out that the new behavior caused some old code to break due to naming inconsistencies. I think the best option here is to adapt to using type = "shift" with negative values, but it definitely threw me for a loop.

# Minimal reproducible example

dat <- data.table(a = 1:3, b = 2:4)

## Using data.table 1.12.2
dat[ , data.table::shift(.SD, 0:1, give.names = TRUE, type = "lead")]
#    a_lag_0 a_lead_1 b_lag_0 b_lead_1
# 1:       1        2       2        3
# 2:       2        3       3        4
# 3:       3       NA       4       NA

## Using data.table 1.11.8
dat[ , data.table::shift(.SD, 0:1, give.names = TRUE, type = "lead")]
#    a_lead_0 a_lead_1 b_lead_0 b_lead_1
# 1:        1        2        2        3
# 2:        2        3        3        4
# 3:        3       NA        4       NA

# Output of sessionInfo()

R-3.5.1; data.table 1.12.2 & 1.11.8

[MichaelChirico]

@jangorecki it's this line of code:

6e60d7b

with specific comment:

lead_0 should be named lag_0 for consistency

I think that makes more sense if there are mixed positive & negative n, seems natural if all(n>=0) to use only one word, thoughts?