Remove SET_OBJECT() usage

[DavisVaughan]

Part of #1933

In vec_is_unspecified(), we no longer check directly against the ATTRIB() pointer when we have an actual <vctrs_unspecified> object. I see no discernible difference between doing that and just using Rf_inherits()

cross::bench_branches({
  library(vctrs)
  is_unspecified <- vctrs:::is_unspecified
  x <- unspecified(1e6)
  bench::mark(is_unspecified(x))
})
#> # A tibble: 2 × 7
#>   branch                expression             min   median `itr/sec` mem_alloc `gc/sec`
#>   <chr>                 <bch:expr>        <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 feature/no-set-object is_unspecified(x)    123ns    205ns  4633428.        0B        0
#> 2 main                  is_unspecified(x)    164ns    205ns  4248652.        0B        0

cross::bench_branches({
  library(vctrs)
  is_unspecified <- vctrs:::is_unspecified
  y <- rep(FALSE, 1e6)
  bench::mark(is_unspecified(y))
})
#> # A tibble: 2 × 7
#>   branch                expression             min   median `itr/sec` mem_alloc `gc/sec`
#>   <chr>                 <bch:expr>        <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 feature/no-set-object is_unspecified(y)    164ns    205ns  4316017.        0B        0
#> 2 main                  is_unspecified(y)    164ns    205ns  4166605.        0B        0

cross::bench_branches({
  library(vctrs)
  is_unspecified <- vctrs:::is_unspecified
  z <- rep(NA, 1e6)
  bench::mark(is_unspecified(z))
})
#> # A tibble: 2 × 7
#>   branch                expression             min   median `itr/sec` mem_alloc `gc/sec`
#>   <chr>                 <bch:expr>        <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 feature/no-set-object is_unspecified(z)    564µs    571µs     1739.        0B        0
#> 2 main                  is_unspecified(z)    564µs    585µs     1706.        0B        0