IVGCVSW-5826 Change weights layout for depthwise to [1,H,W,I*M]

 * This change is necessary because tflite uses a [1,H,W,I*M] format
   and uses the I*M dimension for per axis quantization. Our previous
   layout [M,I,H,W] can't handle the correlating quantization scales.
 * Updates Onnx-, TfLiteParser and TfliteDelegate
 * Updates the CpuRef, CpuAcc and GpuAcc backends
 * Adjusts unit tests
 * Adds test to ensure models with old layout can still be read and
   executed
 * Adds conversion function to previous layout [1,H,W,I*M] --> [M,I,H,W]
   which can be used by backend developers

!android-nn-driver:5553

Signed-off-by: Jan Eilers <jan.eilers@arm.com>
Change-Id: Ifef23368b8c3702cf315a5838d214f7dc13c0152
diff --git a/src/backends/reference/workloads/Decoders.hpp b/src/backends/reference/workloads/Decoders.hpp
index 0b3f360..cd0dc5d 100644
--- a/src/backends/reference/workloads/Decoders.hpp
+++ b/src/backends/reference/workloads/Decoders.hpp
@@ -20,11 +20,7 @@
 
 inline std::unique_ptr<Decoder<float>> MakeSigned32PerAxisDecoder(const TensorInfo& info, const void* data)
 {
-    auto params = armnnUtils::GetPerAxisParams(info);
-    return std::make_unique<ScaledInt32PerAxisDecoder>(
-        static_cast<const int32_t*>(data),
-        params.second,
-        params.first);
+    return std::make_unique<ScaledInt32PerAxisDecoder>(static_cast<const int32_t*>(data), info);
 }
 
 inline std::unique_ptr<Decoder<float>> MakeSigned32Decoder(const TensorInfo& info, const void* data)
@@ -75,10 +71,7 @@
         case armnn::DataType::QuantizedSymm8PerAxis:
         {
             std::pair<unsigned int, std::vector<float>> params = armnnUtils::GetPerAxisParams(info);
-            return std::make_unique<QSymm8PerAxisDecoder>(
-                static_cast<const int8_t*>(data),
-                params.second,
-                params.first);
+            return std::make_unique<QSymm8PerAxisDecoder>(static_cast<const int8_t*>(data), info);
         }
         ARMNN_NO_DEPRECATE_WARN_END
         case DataType::QAsymmS8:
@@ -123,10 +116,7 @@
             if (info.HasPerAxisQuantization())
             {
                 std::pair<unsigned int, std::vector<float>> params = armnnUtils::GetPerAxisParams(info);
-                return std::make_unique<QSymm8PerAxisDecoder>(
-                    static_cast<const int8_t*>(data),
-                    params.second,
-                    params.first);
+                return std::make_unique<QSymm8PerAxisDecoder>(static_cast<const int8_t*>(data), info);
             }
             else
             {