Fix inconsistencies in reporting benchmarking failures

With these changes, all these benchmarks
```cpp
BENCHMARK("Empty benchmark") {};
BENCHMARK("Throwing benchmark") {
    throw "just a plain literal, bleh";
};
BENCHMARK("Asserting benchmark") {
    REQUIRE(1 == 2);
};
BENCHMARK("FAIL'd benchmark") {
    FAIL("This benchmark only fails, nothing else");
};
```

report the respective failure and mark the outer `TEST_CASE` as
failed. Previously, the first two would not fail the `TEST_CASE`,
and the latter two would break xml reporter's formatting, because
`benchmarkFailed`, `benchmarkEnded` etc would not be be called
properly in failure cases.

tests/SelfTest/IntrospectiveTests/InternalBenchmark.tests.cpp

@@ -412,3 +412,24 @@ TEST_CASE("run benchmark", "[benchmark][approvals]") {
 
     CHECK((end - start).count() == 2867251000);
 }
+
+TEST_CASE("Failing benchmarks", "[!benchmark][.approvals]") {
+    SECTION("empty", "Benchmark that has been optimized away (because it is empty)") {
+        BENCHMARK("Empty benchmark") {};
+    }
+    SECTION("throw", "Benchmark that throws an exception") {
+        BENCHMARK("Throwing benchmark") {
+            throw "just a plain literal, bleh";
+        };
+    }
+    SECTION("assert", "Benchmark that asserts inside") {
+        BENCHMARK("Asserting benchmark") {
+            REQUIRE(1 == 2);
+        };
+    }
+    SECTION("fail", "Benchmark that fails inside") {
+        BENCHMARK("FAIL'd benchmark") {
+            FAIL("This benchmark only fails, nothing else");
+        };
+    }
+}