Add execution policy for the tests

[Yaraslaut]

It would be great to be able to run tests in parallel without any additional code, this still should be configurable by the user, maybe as a cli arguments

raw example would be

template <auto refl> 
void run_tests() {
  std::vector<std::thread> pool;
  [: expand(members_of(refl, std::meta::access_context::current())):] >> [&]<auto mem>
  {
      pool.emplace_back(std::thread([&]() {
          [:mem:]();
      }));
   };
   for (auto &t : pool) {
      t.join();
  }
}

[Tsche]

That's an interesting idea, but I'm not sure it is doable.

A lot of things currently assume single-threaded operation outside of tests. For instance assertion counters and coverage will break - I will no longer be able to tell which one of the running tests contributed. Races on coverage counters are mostly okay (after all we're interested in what was reached, not necessarily how often), but this would also cause issues with multiple threads trying to enable/disable coverage tracking concurrently.

Down the line fuzz testing should be supported - libFuzzer already parallelizes by itself. This would have to run outside of a possible multi-threaded mode as well.

[Tsche]

Will address this with the new execution model