Improved ONNX export for relu6 layer

[bas-aarts]

the current ONNX export for relu6 generates a min/max pair (

onnxscript/onnxscript/function_libs/torch_lib/ops/nn.py

Lines 1594 to 1599 in 4cd022b

	@torch_op("aten::relu6", trace_only=True)
	def aten_relu6(self: TReal) -> TReal:
	"""relu6(Tensor self) -> Tensor"""
	six = op.CastLike(op.Constant(value_int=6), self)
	return op.Min(op.Relu(self), six)

)

given that ONNX has a Clip layer, Why not use

    zero = op.CastLike(op.Constant(value_int=0), self)
    six = op.CastLike(op.Constant(value_int=6), self)
    return op.Clip(self, zero, six)

instead?

Sure, the sequence is easy to optimize downstream, but generating the Clip layer here makes a lot of sense.