Use split v1 builder in all onnx ops

src/ngraph/frontend/onnx_import/op/lp_pool.cpp

@@ -45,7 +45,8 @@ namespace ngraph
                     ASSERT_VALID_ARGUMENT(node, p_norm >= 0)
                         << "Only positive (including zero) values are supported for 'p' attribute.";
 
-                    NodeVector slices = ngraph::builder::split(data, channels_count, channel_axis);
+                    NodeVector slices =
+                        ngraph::builder::opset1::split(data, channels_count, channel_axis);
 
                     for (auto& slice : slices)
                     {

src/ngraph/frontend/onnx_import/op/lstm.cpp

@@ -91,7 +91,7 @@ namespace ngraph
                         if (ng_inputs.size() > 3 && !ng_inputs.at(3)->is_null())
                         {
                             auto bias = ng_inputs.at(3);
-                            auto split_bias = builder::split(bias, 2, 1);
+                            auto split_bias = builder::opset1::split(bias, 2, 1);
                             m_map[LSTMInput::LSTM_INPUT_B] = split_bias.at(0) + split_bias.at(1);
                         }
                         else

src/ngraph/frontend/onnx_import/op/pad.cpp

@@ -129,7 +129,7 @@ namespace ngraph
                     {
                         auto axis =
                             default_opset::Constant::create(element::i64, ngraph::Shape{}, {0});
-                        NodeVector padding = builder::split(pads, 2, 0);
+                        NodeVector padding = builder::opset1::split(pads, 2, 0);
 
                         padding_begin =
                             std::make_shared<default_opset::Convert>(padding.at(0), element::i64);