QDQ onnx export

WIP

This MR will be linked with a a MR on aidge_onnx, see aidge_onnx#61.

The goal is to introduce the support of quantized model in aidge_onnx.

Changes

• Add function set to qdq
• Fix binding of append_round_clip and cast_quantizer_ios. Pybind doesn't support reference of smart pointer well as a function argument.
• Minor update on set_scaling function, do not create a producer but instead directly update the Tensor value.
• PTQ no longer convert the calibration dataset to float64, the datatype is reverted before ending the PTQ
• Add function set_scaling_factor_dtype that allow to change the datatype of the scaling factor and the cast operator if it exist.

TODO

• Add set to qdq
• Add unit test on set qdq