table with explicit factor coercion for integer64 in multiple input case

[hcirellu]

In main the default table function is used for multiple inputs if the inputs are not all integer64 or mixed with integer. This leads to correct results, but unexpected ordering for the integer64 inputs.

x <- sample(as.integer64(c(132724613L, -2143220989L, -1L, NA, 1L)), rep=TRUE, size = 1000000)
y <- sample(c(TRUE, FALSE), replace = TRUE, size = length(x))
table(x, y, useNA = "ifany")

main:

#              y
# x              FALSE   TRUE
#   <NA>        100165 100002
#   1            99567  99807
#   132724613   100446 100119
#   -1           99881  99922
#   -2143220989 100129  99962

this PR:

#              y
# x              FALSE   TRUE
#   -2143220989 100129  99962
#   -1           99881  99922
#   1            99567  99807
#   132724613   100446 100119
#   <NA>        100165 100002

microbenchmark::microbenchmark(main=table(x, y, useNA="ifany"), PR=table(x, y, useNA="ifany"))
# Unit: milliseconds
#  expr      min       lq     mean   median       uq      max neval
#  main 115.1219 129.1376 143.6508 133.3782 145.9979 362.4458   100
#    PR  83.2341  91.2439 128.9546  97.7848 111.4765 332.4179   100

This PR requires #255 in order to work properly. The commented test in test-highlevel64.R has to be uncommented.

Closes #236