Add convenience function to write resources_per_worker into a spec

[lilyminium]

Is your feature request related to a problem? Please describe.

As described in #633, I couldn't rescale my gpu workers to account for a lower workload. However, I could modify them locally when I deleted the worker group and re-added it directly with dask-kubernetes (KubeCluster.add_worker_group(custom_spec)). I copied this spec manually, which was mildly inconvenient.

Describe the solution you'd like

It'd be nice to add a BaseDaskKubernetesBackend.make_worker_spec(group_name) function to make this slightly easier.

Describe alternatives you've considered

Additional context