Fix fuzzer range error for scalar parity check.

Reference issue

b/421827351

What does this implement/fix?

The TensorFlow fuzzer found an type overflow error when converting x to a long in the condition. This removes the long cast.

Additional information