add factor and ordered for integer64

[hcirellu]

The base coercion from integer64 to factor does not work properly regarding the sorting of the levels.

Actual behavior:

factor(as.integer64(c(132724613L, -2143220989L, -1L, NA, 1L)))
# [1] 132724613   -2143220989 -1          <NA>        1          
# Levels: 1 132724613 -2143220989 -1
factor(c(132724613L, -2143220989L, -1L, NA, 1L))
# [1] 132724613   -2143220989 -1          <NA>        1          
# Levels: -2143220989 -1 1 132724613

Expected behavior for integer64:

factor(as.integer64(c(132724613L, -2143220989L, -1L, NA, 1L)))
# [1] 132724613   -2143220989 -1          <NA>        1          
# Levels: -2143220989 -1 1 132724613

In order to have table work consistent with base R, this correct factor conversion is necessary.

In addition, the explicit conversion should also be faster for larger vectors.

x <- sample(as.integer64(c(132724613L, -2143220989L, -1L, NA, 1L)), rep=T, size = 1000000)
microbenchmark::microbenchmark(factor(x), base::factor(x))
# Unit: milliseconds
#             expr     min       lq     mean   median       uq      max neval
#        factor(x) 25.5235 28.22665 50.36219 35.63280 41.25975 238.0824   100
#  base::factor(x) 70.0936 78.46005 88.04246 82.59635 89.18355 291.1204   100

[MichaelChirico]

Basically ready to go, just a few dangling comments to make sure the edits to the benchmarking code are WAI.

Feel free to merge at your discretion.