[webgpu] add cases to cover the subgroup matrix path (microsoft#24970)

### Description
<!-- Describe your changes. -->



### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->

Author: xhcao

onnxruntime/test/contrib_ops/matmul_4bits_test.cc

@@ -446,6 +446,8 @@ TEST(MatMulNBits, Float16_Accuracy4) {
   TestMatMulNBitsTyped<MLFloat16, 100, 32, 16, 16, 4>();
   TestMatMulNBitsTyped<MLFloat16, 100, 32, 32, 16, 4>();
   TestMatMulNBitsTyped<MLFloat16, 100, 32, 16, 128, 4>();
+  TestMatMulNBitsTyped<MLFloat16, 100, 64, 32, 32, 4>();
+  TestMatMulNBitsTyped<MLFloat16, 100, 128, 128, 32, 4>();
   TestMatMulNBitsTyped<MLFloat16, 100, 288, 16, 16, 4>();
   TestMatMulNBitsTyped<MLFloat16, 100, 288, 1024, 16, 4>();
   TestMatMulNBitsTyped<MLFloat16, 100, 288, 1024, 128, 4>();

onnxruntime/test/contrib_ops/matmul_8bits_test.cc

@@ -289,6 +289,8 @@ TEST(MatMulNBits, Float16_8b_AccuracyLevel4) {
   constexpr float abs_error = 0.055f;
   constexpr float rel_error = 0.02f;
   TestMatMul8BitsTyped<MLFloat16, 2, 4, 32, 16, 4>(abs_error, rel_error);
+  TestMatMul8BitsTyped<MLFloat16, 100, 64, 32, 32, 4>(abs_error, rel_error);
+  TestMatMul8BitsTyped<MLFloat16, 100, 128, 128, 32, 4>(abs_error, rel_error);
   TestMatMul8BitsTyped<MLFloat16, 199, 40, 576, 32, 4>(abs_error, rel_error);
 }
 #endif