Match what is in the SDM paper

[jsadler2]

Xiaowei has a slightly different, simpler version of the RGCN in the SDM paper. I think this is more understandable and it'd be good to change to it:

[SimonTopp]

I believe this is implemented in the PyTorch version by @jdiaz4302. @jsadler2, do you think we should implement it in the TF version as well still?

[jdiaz4302]

Yep, that's what going on here: https://github.com/USGS-R/river-dl/blob/main/river_dl/torch_models.py#L164-L165

That # note: self.A @ q_t could be taken out, I believe that's a development relic/reminder that I didn't remove; it doesn't add anything/act as an informative comment.