[QNN EP] Re-enable several disabled QNN-EP UTs

### Description
1. Re-enable UTs which passed 2.30
2. Fix conv and resize UTs
   - Make conv's weight as initializer to let graph.NumberOfNodes()
     match ep_nodes, which should be 1.
   - Update resize UT because "round_prefer_floor" is no longer
     supported in QNN SDK since 2.21.

### Motivation and Context
1. Make the UT of QNN EP pass as much as possible to improve the test
   coverage.

Author: kuanyul-quic

onnxruntime/test/providers/qnn/conv_test.cc

@@ -1519,9 +1519,10 @@ TEST_F(QnnHTPBackendTests, Conv3D_U16S8S32_PerChannel2) {
 // Expected val: 87.354057312011719
 // QNN QDQ val: 0 (err 87.354057312011719)
 // CPU QDQ val: 87.3583984375 (err 0.00434112548828125)
-TEST_F(QnnHTPBackendTests, DISABLED_ConvU16S16S32_DynamicBias) {
+// Issue fixed in 2.30
+TEST_F(QnnHTPBackendTests, ConvU16S16S32_DynamicBias) {
   TestInputDef<float> input_def({1, 2, 5, 5}, false, GetFloatDataInRange(-10.0f, 10.0f, 50));
-  TestInputDef<float> weight_def({1, 2, 3, 3}, false, GetFloatDataInRange(-1.0f, 5.0f, 18));
+  TestInputDef<float> weight_def({1, 2, 3, 3}, true, GetFloatDataInRange(-1.0f, 5.0f, 18));
   RunHTPConvOpTest<uint16_t, int16_t>("Conv",
                                       input_def,                                   // Input
                                       weight_def.OverrideValueRange(-5.0f, 5.0f),  // Weights (symmetric quant range)
@@ -1537,9 +1538,10 @@ TEST_F(QnnHTPBackendTests, DISABLED_ConvU16S16S32_DynamicBias) {
 
 // Tests 16-bit QDQ Conv with dynamic weights and bias (uses QNN's DepthwiseConv2d)
 // TODO(adrianlizarraga): FAIL: Failed to finalize QNN graph. Error code 1002
-TEST_F(QnnHTPBackendTests, DISABLED_DepthwiseConvU16S16S32_DynamicBias) {
+// Issue fixed in 2.30
+TEST_F(QnnHTPBackendTests, DepthwiseConvU16S16S32_DynamicBias) {
   TestInputDef<float> input_def({1, 1, 5, 5}, false, GetFloatDataInRange(-10.0f, 10.0f, 25));
-  TestInputDef<float> weight_def({1, 1, 3, 3}, false, GetFloatDataInRange(-1.0f, 5.0f, 9));
+  TestInputDef<float> weight_def({1, 1, 3, 3}, true, GetFloatDataInRange(-1.0f, 5.0f, 9));
   RunHTPConvOpTest<uint16_t, int16_t>("Conv",
                                       input_def,                                   // Input
                                       weight_def.OverrideValueRange(-5.0f, 5.0f),  // Weights (symmetric quant range)
@@ -1559,9 +1561,10 @@ TEST_F(QnnHTPBackendTests, DISABLED_DepthwiseConvU16S16S32_DynamicBias) {
 // Expected val: 85.354057312011719
 // QNN QDQ val: 0 (err 85.354057312011719)
 // CPU QDQ val: 85.358139038085938 (err 0.00408172607421875)
-TEST_F(QnnHTPBackendTests, DISABLED_ConvU16S16S32_NoBias) {
+// Issue fixed in 2.30
+TEST_F(QnnHTPBackendTests, ConvU16S16S32_NoBias) {
   TestInputDef<float> input_def({1, 2, 5, 5}, false, GetFloatDataInRange(-10.0f, 10.0f, 50));
-  TestInputDef<float> weight_def({1, 2, 3, 3}, false, GetFloatDataInRange(-1.0f, 5.0f, 18));
+  TestInputDef<float> weight_def({1, 2, 3, 3}, true, GetFloatDataInRange(-1.0f, 5.0f, 18));
   RunHTPConvOpTest<uint16_t, int16_t>("Conv",
                                       input_def,                                   // Input
                                       weight_def.OverrideValueRange(-5.0f, 5.0f),  // Weights (symmetric quant range)
@@ -1577,12 +1580,13 @@ TEST_F(QnnHTPBackendTests, DISABLED_ConvU16S16S32_NoBias) {
 
 // Tests 16-bit QDQ Conv with dynamic weights and no bias (uses QNN's DepthWiseConv2d)
 // TODO(adrianlizarraga): FAIL: Failed to finalize QNN graph. Error code 1002
-TEST_F(QnnHTPBackendTests, DISABLED_DepthwiseConvU16S16S32_NoBias) {
+// Issue fixed in 2.30
+TEST_F(QnnHTPBackendTests, DepthwiseConvU16S16S32_NoBias) {
   std::vector<float> input_data = GetFloatDataInRange(-10.0f, 10.0f, 25);
   std::vector<float> weight_data = GetFloatDataInRange(-10.0f, 10.0f, 9);
   RunHTPConvOpTest<uint16_t, int16_t>("Conv",
                                       TestInputDef<float>({1, 1, 5, 5}, false, input_data),   // Input
-                                      TestInputDef<float>({1, 1, 3, 3}, false, weight_data),  // Weights
+                                      TestInputDef<float>({1, 1, 3, 3}, true, weight_data),   // Weights
                                       TestInputDef<float>(),                                  // Bias
                                       {1, 1},                                                 // Strides
                                       {0, 0, 0, 0},                                           // Pads

onnxruntime/test/providers/qnn/resize_test.cc

@@ -456,11 +456,12 @@ TEST_F(QnnHTPBackendTests, ResizeU8_2xNearestAsymmetricRoundPreferFloor_Unsuppor
 // ORT CPU EP (f32 model) outputs: -10.0000000 -10.0000000 -3.33333349 -3.33333349 -3.33333349 -3.33333349 -10.00 ...
 // ORT CPU EP (qdq model) outputs: -9.96078491 -9.96078491 -3.29411769 -3.29411769 -3.29411769 -3.29411769 -9.961 ...
 // ORT QNN EP (qdq model) outputs: -9.96078491 -9.96078491 -9.96078491 -3.37254906 -3.37254906 -3.37254906 -9.961 ...
-TEST_F(QnnHTPBackendTests, DISABLED_ResizeU8_3xNearestAsymmetricRoundPreferFloor) {
+// UPDATE: "round_prefer_floor" no longer supported in QNN SDK 2.21 (supported in QNN SDK 2.19)
+TEST_F(QnnHTPBackendTests, ResizeU8_3xNearestAsymmetricRoundPreferFloor_Unsupported) {
   std::vector<float> input_data = GetFloatDataInRange(-10.0f, 10.0f, 4);
   RunQDQResizeOpTest<uint8_t>(TestInputDef<float>({1, 1, 2, 2}, false, input_data),
                               {1, 1, 6, 6}, "nearest", "asymmetric", "round_prefer_floor",
-                              ExpectedEPNodeAssignment::All);
+                              ExpectedEPNodeAssignment::None);  // No longer supported as of QNN SDK 2.21
 }
 
 // Test 0.5x QDQ Resize mode: "nearest", coordinate_transformation_mode: "asymmetric", nearest_mode: "floor"

onnxruntime/test/providers/qnn/simple_op_htp_test.cc

@@ -297,7 +297,8 @@ TEST_F(QnnHTPBackendTests, UnaryOp_Elu) {
 // Expected val: -0.99751651287078857
 // QNN QDQ val: 6.2726154327392578 (err 7.2701320648193359)
 // CPU QDQ val: -0.99753034114837646 (err 1.3828277587890625e-05)
-TEST_F(QnnHTPBackendTests, DISABLE_UnaryOp_Elu_U16) {
+// Issue fixed in 2.30
+TEST_F(QnnHTPBackendTests, UnaryOp_Elu_U16) {
   RunQDQOpTest<uint16_t>("Elu",
                          {TestInputDef<float>({1, 2, 3}, false, GetFloatDataInRange(-10.0f, 10.0f, 6))},
                          {},