Fix big endian handling in totypes (#209)

Previously the dtype was viewed, but an actual dtype conversion needs to
happen.

Signed-off-by: Justin Chu <justinchuby@users.noreply.github.com>

Author: justinchuby

src/onnx_ir/_core.py

@@ -520,7 +520,7 @@ def tobytes(self) -> bytes:
         else:
             assert self.dtype.itemsize == array.itemsize, "Bug: The itemsize should match"
         if not _IS_LITTLE_ENDIAN:
-            array = array.view(array.dtype.newbyteorder("<"))
+            array = array.astype(array.dtype.newbyteorder("<"))
         return array.tobytes()
 
 
@@ -1110,7 +1110,7 @@ def tobytes(self) -> bytes:
         """
         array = self.numpy_packed()
         if not _IS_LITTLE_ENDIAN:
-            array = array.view(array.dtype.newbyteorder("<"))
+            array = array.astype(array.dtype.newbyteorder("<"))
         return array.tobytes()