Filter NA in fancy_breaks() big-number abbreviation check

[SAY-5]

Fixes #1259.

fancy_breaks() errored with "missing value where TRUE/FALSE needed" when a numeric column contained NA and the values were large enough to trigger the big.num.abbr abbreviation path (values >= 1e6). This surfaced when formatting popups in view mode, even when the NA column was not the one used for fill.

The abbreviation check all(vec_fin - floor(vec_fin/(10^o))*(10^o) < 1) returns NA when vec_fin contains NA, so if (...) throws. vec_fin was built by filtering infinite values but not NA. This filters NA too, in both places vec_fin is derived.

tmap:::fancy_breaks(c(1e6, NA, 3e6))
# before: Error: missing value where TRUE/FALSE needed
# after:  "1 mln" "NA mln" "3 mln"

Added a regression test in tests/testthat/test-fancy_breaks.R and a NEWS entry.

[olivroy]

Hello @SAY-5, it seems like the new test you added fails.

[SAY-5]

Good catch, the test was too strict. fancy_breaks returns a formatted vector with an align attribute from format(), so I swapped expect_identical for expect_equal(..., ignore_attr = TRUE) which only checks the values.