ENH: Add functionality to set seed

[NickleDave]

We should add functionality to set the seed for torch

This came up because @VinCheetah is having a bit of trouble replicating results from the 2022 eLife TweetyNet paper

A separate question is whether we want to "set the seed for everything", I see some libraries do something like this (from https://gist.github.com/gatheluck/c57e2a40e3122028ceaecc3cb0d152ac)

import os
import random
import numpy as np
import torch

def set_all_seeds(seed):
  random.seed(seed)
  os.environ('PYTHONHASHSEED') = str(seed)
  np.random.seed(seed)
  torch.manual_seed(seed)
  torch.cuda.manual_seed(seed)
  torch.backends.cudnn.deterministic = True

Earlier versions of lightning similarly had a function seed_everything (that AFAICT has since migrated to fabric): https://pytorch-lightning.readthedocs.io/en/1.7.7/api/pytorch_lightning.utilities.seed.html#pytorch_lightning.utilities.seed.seed_everything

but just doing a bare minimum of setting the torch seed will let us control e.g. random initialization of model parameters