Use memcpy() of standard C library

[Watson1978]

Ruby has its own implementation of ruby_nonempty_memcpy(), which is used when memcpy() is called.

static inline void *
ruby_nonempty_memcpy(void *dest, const void *src, size_t n)
{
    if (n) {
        return memcpy(dest, src, n);
    }
    else {
        return dest;
    }
}
RBIMPL_SYMBOL_EXPORT_END()
#undef memcpy
#define memcpy ruby_nonempty_memcpy

https://github.com/ruby/ruby/blob/master/include/ruby/internal/memory.h

It has an unnecessary if statement and it has some overhead.

Similar: #735

−	before	after	result
Oj.dump (macOS)	7.839k	8.000k	1.021x
Oj.dump (Linux)	9.140k	10.043k	1.099x

Environment

• macOS
  • macOS 12.1
  • Apple M1 Max
  • Apple clang version 13.0.0 (clang-1300.0.29.30)
  • Ruby 3.1.0
• Linux
  • Zorin OS 16
  • AMD Ryzen 7 5700G
  • gcc version 11.1.0
  • Ruby 3.1.0

macOS

Before

Warming up --------------------------------------
             Oj.dump   783.000  i/100ms
Calculating -------------------------------------
             Oj.dump      7.839k (± 0.9%) i/s -    118.233k in  15.084888s

After

Warming up --------------------------------------
             Oj.dump   803.000  i/100ms
Calculating -------------------------------------
             Oj.dump      8.000k (± 0.8%) i/s -    120.450k in  15.057950s

Linux

Before

Warming up --------------------------------------
             Oj.dump   887.000  i/100ms
Calculating -------------------------------------
             Oj.dump      9.140k (± 0.9%) i/s -    137.485k in  15.043569s

After

Warming up --------------------------------------
             Oj.dump   972.000  i/100ms
Calculating -------------------------------------
             Oj.dump     10.043k (± 0.9%) i/s -    150.660k in  15.002236s

Test code

require 'benchmark/ips'
require 'oj'

data = (0..10000).to_a

Benchmark.ips do |x|
  x.time = 15

  x.report('Oj.dump') { Oj.dump(data, mode: :compat) }
end

[Watson1978]

Here is stacktrace with before changing.