sky logs and sky queue should be accessible on stopped instances

[csaroff]

When running sky logs on a stopped instance, I receive the following message:

$ sky logs sky-bench-hydra-3eps-tk2-1
sky.exceptions.ClusterNotUpError: Tailing logs: skipped for cluster 'sky-bench-hydra-3eps-tk2-1' (status: STOPPED). It is only allowed for UP clusters.

Is it possible to have each cluster stream its logs to a cloud storage bucket and have sky logs reference that instead? It's a bit tedious(and potentially expensive) to restart clusters just to pull down the logs after a job completes.