Remove intermediate casting in product

Created by: jmitrevs

This removes the intermediate casting of scale*x to a res_T type in a BatchNormalization, so casting a res_t is only of the final scale*x + bias. I did this by introducing Product_nocast and mult_nocast, which multiply and return in the full precision. That seemed easier and less error-prone than manually determining the resulting type in python, and I was hesitant to always make mult and Product return the full precision, though that would be straightforward.

I did confirm that the simple test model is synthesizable with Vivado_HLS 2019.2.

This patch does fix some issues I saw working with low bit values on QONNX.