⚡ Skip replacing NAs during mapping when appropriate

[teunbrand]

This PR replaces #5032.

Briefly, it skips the NA replacment step if there are no NAs present.
Moreover vec_assign() is faster (and I presume more type stable) than ifelse().
I think it is more of an 'empirical' determination than relying on some palette attribute as is proposed in the linked PR.

Benchmarking this PR, note 4 continuous scales (x, y, size, colour):

devtools::load_all("~/packages/ggplot2/")
#> ℹ Loading ggplot2

p <- ggplot(diamonds, aes(carat, price, colour = depth, size = table)) +
  geom_point()

bench::mark(ggplot_build(p), min_iterations = 10)
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 1 × 6
#>   expression           min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>      <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 ggplot_build(p)   31.4ms   34.1ms      15.3    50.4MB     24.5

^{Created on 2025-02-24 with reprex v2.1.1}

The same code on the main branch gives me:

#> # A tibble: 1 × 6
#>   expression           min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>      <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 ggplot_build(p)   39.9ms   41.1ms      15.4    66.5MB     32.4

^{Created on 2025-02-24 with reprex v2.1.1}

[teunbrand]

Second benchmark, this time with NAs present:

devtools::load_all("~/packages/ggplot2/")
#> ℹ Loading ggplot2

df <- diamonds
set.seed(42)
df$carat[sample(nrow(df), 5)] <- NA
df$price[sample(nrow(df), 50)] <- NA
df$depth[sample(nrow(df), 500)] <- NA
df$table[sample(nrow(df), 5000)] <- NA

p <- ggplot(df, aes(carat, price, colour = depth, size = table)) +
  geom_point()

bench::mark(ggplot_build(p), min_iterations = 10)
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 1 × 6
#>   expression           min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>      <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 ggplot_build(p)   32.4ms     36ms      14.7    53.9MB     28.0

^{Created on 2025-02-24 with reprex v2.1.1}

Same for main:

#> # A tibble: 1 × 6
#>   expression           min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>      <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 ggplot_build(p)   40.6ms   43.4ms      13.4    68.7MB     28.2

[thomasp85]

LGTM