cl.locksmith.cluster is flakey

[ajeddeloh]

It looks like there's a race with the cl.locksmith.cluster test where the system stops sshd too fast and causes the ssh command to fail which then causes the test to fail.

example:

--- FAIL: cl.locksmith.cluster (476.76s)
        locksmith.go:169: [0] failed to run "sudo systemctl stop sshd.socket && sudo locksmithctl send-need-reboot": output: "" status: "ssh: handshake failed: EOF" [1] failed to run "sudo systemctl stop sshd.socket && sudo locksmithctl send-need-reboot": output: "" status: "ssh: handshake failed: EOF" [2] ssh unreachable: context canceled