[BUG] DoNotOptimize const & overload chosen for non-const lvalue references

[eseiler]

Describe the bug
After df9a99d, const & overloads are chosen where previously the & overload was chosen. This leads to deprecation warnings where IMO there should be none. const & takes precedence over the new && overload.

Godbolt: https://godbolt.org/z/5s6sKT7dc

Click to show code from godbolt example

#include <benchmark/benchmark.h>

#include <algorithm>
#include <vector>

static void test(benchmark::State & state)
{
    std::vector<int> vec(100);

    for (auto _ : state)
    {
        std::fill(vec.begin(), vec.end(), 1);
        benchmark::DoNotOptimize(vec);
    }
}
BENCHMARK(test);

BENCHMARK_MAIN();

<source>: In function 'void test(benchmark::State&)':
<source>:13:33: error: 'typename std::enable_if<((! std::is_trivially_copyable<_Tp>::value) || (sizeof (Tp) > sizeof (Tp*)))>::type benchmark::DoNotOptimize(const Tp&) [with Tp = std::vector<int>; typename std::enable_if<((! std::is_trivially_copyable<_Tp>::value) || (sizeof (Tp) > sizeof (Tp*)))>::type = void]' is deprecated: The const-ref version of this method can permit undesired compiler optimizations in benchmarks [-Werror=deprecated-declarations]
   13 |         benchmark::DoNotOptimize(vec);
      |         ~~~~~~~~~~~~~~~~~~~~~~~~^~~~~
In file included from <source>:1:
/opt/compiler-explorer/libs/google-benchmark/trunk/include/benchmark/benchmark.h:502:5: note: declared here
  502 |     DoNotOptimize(Tp const& value) {
      |     ^~~~~~~~~~~~~
cc1plus: all warnings being treated as errors

System
Which OS, compiler, and compiler version are you using:

• OS: Ubuntu 22.04.2 LTS
• Compiler and version: gcc version 12.3.0 (locally), 13.1.0 (godbolt)

To reproduce
See bug description.

Expected behavior
IMO, it should still work. There probably needs to be both & and && overloads. It does work if I add additional & overloads.

[eseiler]

If this is confirmed to be a bug, I could offer to provide a PR with the code changes and test cases.

[LebedevRI]

I'm pretty sure #1608 should have only affected rvalues, so this does seem like a bug.

[LebedevRI]

Hm, i'm confused. DoNotOptimize const & overload change happened much earlier than df9a99d,
it was already present in v1.8.0, yet i concur that i'm also only now started getting those warnings, in v1.8.1:
https://github.com/darktable-org/rawspeed/actions/runs/5454117056/jobs/9923812487?pr=494

@dmah42 so i guess that means the original #1493 didn't prohibit all the cases?

[LebedevRI]

Ok, no, let me backtrack, those are false-positives, this is a bug. @eseiler patches most welcomed!
Looks like we'll be getting .2 soon? (: