fill() should handle NaN

[ArtemSokolov]

Dear tidyverse developers,

In the current implementation, fill() ignores NaN. My current workaround has been a (rather strange-looking) use of replace_na() to replace NaN with NA before passing the values to fill():

X <- tibble::tibble(x = c(1,0,3,0),
                    y = c(2,0,2,0),
                    z = x/y)

tidyr::fill(X, z)
#  # A tibble: 4 x 3
#        x     y     z
#    <dbl> <dbl> <dbl>
#  1     1     2   0.5
#  2     0     0 NaN
#  3     3     2   1.5
#  4     0     0 NaN

tidyr::replace_na(X, list(z=NA)) %>%
    tidyr::fill(z)
# # A tibble: 4 x 3
 #       x     y     z
 #   <dbl> <dbl> <dbl>
 # 1     1     2   0.5
 # 2     0     0   0.5
 # 3     3     2   1.5
 # 4     0     0   1.5

I believe it would be warranted to have fill() treat NaN and NA equivalently, because 1) it is already done so by replace_na(), and 2) is.na(NaN) is TRUE.