The integration tests on self-hosted runners were consistently failing due to insufficient disk space during Juju bootstrap. While debugging with tmate, I found that the default rawfile-storage path under /var/snap/k8s/common/rawfile-storage had only ~19 GiB available, whereas the tests require more than 20 GiB. As a result, K8s was never ready to bootstrap with Juju.

(One of the failed CI runs)[https://github.com/canonical/temporal-admin-k8s-operator/actions/runs/17557796899/job/49866784457]

Steps to Reproduce
1. Run integration tests on a self-hosted runner using Canonical k8s.
2. Observe bootstrap failures due to insufficient space under /var/snap/k8s/common/rawfile-storage.

Proposed Change
• Add a step after Canonical k8s setup to:
• Move rawfile-storage from the root disk to /mnt.
• Symlink it back to /var/snap/k8s/common/rawfile-storage.