Loading
nixos/ollama: switch `loadModels` script to use `parallel --tag`
This should theoretically be (slightly) faster, and will prepend the relevant argument/model before every stdout/stderr message, allowing easy identification of which model is causing an error (eg due to a typo).