A Benchmark Comparison of Imitation Learning-based Control Policies for Autonomous Racing