Arm backend: Fix off by one bug in print from arm_executor_runner (pytorch#14120)

The arm executor runner has ET_NUM_INFERENCES set to 1 by default but
the print after run stated it has done 2 runs.

The for loop doing the execution started on 1 and checked with <=
causing the num_inferences to tick over by one.

Signed-off-by: [email protected]

Author: perheld

examples/arm/executor_runner/arm_executor_runner.cpp

@@ -930,7 +930,7 @@ void run_model(RunnerContext& ctx, const void* model_pte) {
 
   int n = 0;
   StartMeasurements();
-  for (n = 1; n <= num_inferences; n++) {
+  for (n = 0; n < num_inferences; n++) {
     // Run the model.
     status = ctx.method.value()->execute();
     if (status != Error::Ok) {