Remove Caffe2 from resnet bs=128 perf testing for now

The same OOM problem impacting pytorch hits caffe2 as well, likely
because they share an execution engine.

Author: David Goodwin

qa/L0_perf_resnet/test.sh

@@ -123,7 +123,7 @@ done
 STATIC_BATCH=128
 DYNAMIC_BATCH=1
 INSTANCE_CNT=2
-MODEL_NAMES="${TRT_MODEL_NAME} ${TF_MODEL_NAME} ${NETDEF_MODEL_NAME}"
+MODEL_NAMES="${TRT_MODEL_NAME} ${TF_MODEL_NAME}"
 OPTIMIZED_MODEL_NAMES="${TFTRT_MODEL_NAME}"
 
 # Create the TensorRT plan from Caffe model