Reduce compilation costs of benchmarks

We replaced some simple std::algorithm usage by loops, and reduced
header inclusion.

tests/SelfTest/IntrospectiveTests/InternalBenchmark.tests.cpp

@@ -22,6 +22,8 @@
 #include <catch2/benchmark/detail/catch_benchmark_function.hpp>
 #include <catch2/benchmark/detail/catch_estimate_clock.hpp>
 
+#include <numeric>
+
 namespace {
     struct manual_clock {
     public: