[SPARK-35315][TESTS] Keep benchmark result consistent between spark-submit and SBT

### What changes were proposed in this pull request?

Set `IS_TESTING` to true in `BenchmarkBase`, before running benchmarks.

### Why are the changes needed?

Currently benchmark can be done via 2 ways: `spark-submit`, or SBT command. However in the former Spark will miss some properties such as `IS_TESTING`, which is necessary to turn on/off certain behavior like codegen (`spark.sql.codegen.factoryMode`). Therefore, the result could differ between the two. In addition, the benchmark GitHub workflow is using the spark-submit approach.

### Does this PR introduce _any_ user-facing change?

No

### How was this patch tested?

N/A

Closes #32440 from sunchao/SPARK-35315.

Authored-by: Chao Sun <[email protected]>
Signed-off-by: Yuming Wang <[email protected]>

Author: sunchao

core/src/test/scala/org/apache/spark/benchmark/BenchmarkBase.scala

@@ -19,6 +19,8 @@ package org.apache.spark.benchmark
 
 import java.io.{File, FileOutputStream, OutputStream}
 
+import org.apache.spark.internal.config.Tests.IS_TESTING
+
 /**
  * A base class for generate benchmark results to a file.
  * For JDK9+, JDK major version number is added to the file names to distinguish the results.
@@ -42,6 +44,10 @@ abstract class BenchmarkBase {
   }
 
   def main(args: Array[String]): Unit = {
+    // turning this on so the behavior between running benchmark via `spark-submit` or SBT will
+    // be consistent, also allow users to turn on/off certain behavior such as
+    // `spark.sql.codegen.factoryMode`
+    System.setProperty(IS_TESTING.key, "true")
     val regenerateBenchmarkFiles: Boolean = System.getenv("SPARK_GENERATE_BENCHMARK_FILES") == "1"
     if (regenerateBenchmarkFiles) {
       val version = System.getProperty("java.version").split("\\D+")(0).toInt