[onnx] Fix onnx.cast cases between int32 and int64 (#2982)

2 modifications:
1. torch.int64 is enum 4 in TORCH_DTYPE_TO_INT
2. add int32 support

lib/Conversion/TorchOnnxToTorch/DefaultDomainAtoF.cpp

@@ -43,8 +43,10 @@ static int64_t onnxDtypeIntToTorchDtypeInt(int64_t dtypeIntOnnx) {
     switch (dtypeIntOnnx) {
     case 1:
       return 6; // float
+    case 6:
+      return 3; // int32
     case 7:
-      return 5; // int64
+      return 4; // int64
     case 9:
       return 11; // bool
     case 10:

projects/pt1/e2e_testing/xfail_sets.py

@@ -1883,10 +1883,7 @@ ONNX_XFAIL_SET = {
     "BucketizeTensorOutInt32RightModule_basic",
     "ElementwiseToDtypeI64ToI8Module_basic",
     "ElementwiseToDtypeI64ToUI8Module_basic",
-    "HBC_basic",
     "QuantizedMLP_basic",
-    "TypeConversionI1ToI32Module_basic",
-    "TypeConversionI64ToI32Module_basic",
 
     # Failure - onnx_lowering: onnx.Clip
     "NormalizeModule_basic",