add onnx model & dpp memory leak check and use const input data

mindspore/lite/src/runtime/kernel/arm/base/detection_post_process_base.cc

@@ -133,6 +133,16 @@ void DetectionPostProcessBaseCPUKernel::FreeAllocatedBuffer() {
     context_->allocator->Free(params_->selected_);
     params_->selected_ = nullptr;
   }
+  if (desc_.data_type == kNumberTypeInt8) {
+    if (input_boxes_ != nullptr) {
+      context_->allocator->Free(input_boxes_);
+      input_boxes_ = nullptr;
+    }
+    if (input_scores_ != nullptr) {
+      context_->allocator->Free(input_scores_);
+      input_scores_ = nullptr;
+    }
+  }
 }
 
 int DetectionPostProcessBaseCPUKernel::Run() {

mindspore/lite/src/runtime/kernel/arm/fp32/detection_post_process_fp32.cc

@@ -33,8 +33,8 @@ int DetectionPostProcessCPUKernel::GetInputData() {
     MS_LOG(ERROR) << "Input data type error";
     return RET_ERROR;
   }
-  input_boxes_ = reinterpret_cast<float *>(in_tensors_.at(0)->MutableData());
-  input_scores_ = reinterpret_cast<float *>(in_tensors_.at(1)->MutableData());
+  input_boxes_ = reinterpret_cast<float *>(in_tensors_.at(0)->data_c());
+  input_scores_ = reinterpret_cast<float *>(in_tensors_.at(1)->data_c());
   return RET_OK;
 }
 

mindspore/lite/src/runtime/kernel/arm/int8/detection_post_process_int8.cc

@@ -50,7 +50,7 @@ int DequantizeInt8ToFp32Run(void *cdata, int task_id) {
 }
 
 int DetectionPostProcessInt8CPUKernel::Dequantize(lite::Tensor *tensor, float **data) {
-  data_int8_ = reinterpret_cast<int8_t *>(tensor->MutableData());
+  data_int8_ = reinterpret_cast<int8_t *>(tensor->data_c());
   *data = reinterpret_cast<float *>(context_->allocator->Malloc(tensor->ElementsNum() * sizeof(float)));
   if (*data == nullptr) {
     MS_LOG(ERROR) << "Malloc data failed.";

mindspore/lite/test/models_with_several_inputs_or_without_outputs.cfg

@@ -24,3 +24,4 @@ efficientnet_lite3_int8_2.tflite
 efficientnet_lite4_int8_2.tflite
 mtk_transformer_encoder.tflite
 mtk_transformer_decoder_joint.tflite
+ml_ei_facedetection.onnx