Update Dockerfile for nvidia runtime

[nlosg]

Setting the runtime to default to nvidia as these are GPU nodes and will likely run GPU-accelerated workloads only. I'm happy to have this as optional to avoid confusion but I think it's beneficial for users to decide for themselves how to manage the configuration.

What

Creates a K3s configuration file setting the default runtime to nvidia. This configuration is passed at cluster bringup when k3s agent is run.

Why

I don't see any issue for GPU specific nodes and it avoids the overhead of defining the runtime in every YAML definition.

Implications

N/A

[nlosg]

@all-contributors please add @nlosg for code and ideas and doc.

[allcontributors[bot]]

@nlosg

I've put up a pull request to add @nlosg! :tada: