[libc++] Optimizations for algorithms on flat containers

[ldionne]

We could do something like this:

template <class Iterator, class Pred>
  requires same_as<Pred, _Compare> &&
           (__desugars_to_v<__less_tag, _Pred, _Reference, _Reference> ||
            __desugars_to_v<__greater_tag, _Pred, _Reference, _Reference> || ...)
bool is_sorted(__ra_iterator<flat_set<_ValueType, _Compare, _Container>, typename _Container::iterator> first,
               __ra_iterator<flat_set<_ValueType, _Compare, _Container>, typename _Container::iterator> last,
               Pred pred) {
  return true;
}

Other things we probably want to do if we go down that route:

• We might want a typedef for __flat_set_iterator<...> which expands to __ra_iterator?
• We might want to have some sort of trait instead of matching flat_set directly, maybe something like __is_sorted_with_respect_to<Iterator, Predicate> that we can then specialize for any container that is sorted as an invariant?
• In std::flat_set::insert(Iter, Iter) & friends, we could add std::sorted_unique_t when we see that Iter is a fellow flat_set iterator (or any kind of sorted iterator, really).