Implement new_vctr() in C

[DavisVaughan]

I'm hoping to move some of the list-of code to C, since I am using that more internally in tidyr, but that requires moving new_vctr() to C. So this PR does that.

A few notes:

• I went the same route as new_data_frame() and used .External(), which seemed best for working with ...
• Like new_data_frame(), supplying names = as an attribute now overrides existing names
• It is currently impossible to provide class = in ... due to the class argument, but I check that in case we ever make these dots dynamic
• This also required moving names_repair_missing() to C, but that was straightforward
• The actual "setting" of attributes at the end is a little clunky because we don't have access to Rf_shallow_duplicate_attr(), which creates an ALTREP wrapper around an object if you plan on just updating its attributes. That makes setting attributes extremely lightweight and fast. That is available through attributes<-, which is what vec_set_attributes() does on R >=3.6, so we call back to R to be able to call that.

We have tried to keep <vctrs_vctr> out of the main codebase, and I think this still accomplishes that. There is nothing here that couldn't be moved to a standalone package - which I think is the right way to think about where C code can interact with the vctr type.

Here is a benchmark showing fairly minor improvements in performance. I think this will be helpful in the long run, since new_vctr() will be called a lot (like in a reduction of vec_ptype2() calls where the result of each call to vec_ptype2() calls new_vctr(), which is what list-of does).

library(vctrs)

x <- 1:1e6 + 0L

bench::mark(vctr = new_vctr(x), iterations = 1000000)
# CRAN
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 vctr         4.33µs   4.97µs   189041.    29.8KB     15.9

# This PR
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 vctr         2.72µs   3.18µs   301434.    6.41KB     24.4

bench::mark(vctr = new_vctr(x, class = "foo"), iterations = 1000000)
# CRAN
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 vctr         4.09µs   4.61µs   206191.        0B     16.5

# This PR
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 vctr         2.82µs   3.29µs   292022.        0B     26.0

bench::mark(vctr = new_vctr(x, class = "foo", inherit_base_type = TRUE), iterations = 1000000)
# CRAN
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 vctr          4.2µs   4.77µs   198925.        0B     16.3

# This PR
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 vctr         3.07µs   3.57µs   269074.        0B     25.0

bench::mark(vctr = new_vctr(x, a = 1, b = x, class = "foo", inherit_base_type = TRUE), iterations = 1000000)
# CRAN
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 vctr         4.47µs   5.24µs   182313.        0B     17.5

# This PR
#> # A tibble: 1 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 vctr          3.6µs   4.11µs   234050.        0B     25.3

^{Created on 2021-11-17 by the reprex package (v2.0.1)}