Multiplier limit for conv1d to take advantage of pruning

Created by: benjaminkreis

Add new function to nnet_conv.h called compute_multiplier_limit that computes the multiplier limit to take advantage of padding and pruning.

Discussed in #56 (closed)

Perhaps the one other thing to note is that zero weights are detected with

if( weights[index_weight] > 1e-20 || weights[index_weight] < -1e-20 ){

This is different from how we have done it with the fully connected layers. In that case we check if the weight ==0 at full precision in python here. Here we don't use the boolean expression, and we look at the weight at ap_fixed precision.