Implement spend_cpu_cycles in assembly

The compiler was unrolling it previously, so we're doing assembly now for consistency.

src/config.hpp

@@ -79,8 +79,8 @@ struct SubscriptionConfig {
  */
 struct BenchmarkConfig {
   /// CPU-intensive loop iterations for /ok requests in resolve stage
-  /// 0 = health check only, 7000 = default benchmark load (650ns, 1M req/s)
-  int ok_resolve_iterations = 7000;
+  /// 0 = health check only, 4000 = default benchmark load (740ns, 1M req/s)
+  int ok_resolve_iterations = 0;
 };
 
 /**