GarNet and GarNetStack in config.py

Created by: yiiyama

I was asked by a Ph.D. student at Imperial College London why hls4ml config doesn't work with GarNet. Found out that graph_layers support isn't there in config.py - which is of course the case because I didn't do it (hls4ml config wasn't really a thing when I was working on GarNet). This PR adds support to GarNet and GarNetStack in config.py and gives some default precision values for their parameters. These default values are however most likely not optimal, because in config.py we don't deal with tensor shapes and therefore we don't know the number of vertices that goes into the layer, which is a critical input when computing optimal precisions. So the lines that gets generated with hls4ml config are there mostly for documentation purpose to say "hey these are the configurable parameters", and users are actually better off commenting them out when actually doing a conversion (proper default precisions will be calculated using tensor shapes), unless they really have specific precision values they want to use. Is this something to be documented somewhere? (add a comment line in the generated config?)