CI: staging-test-with-rebase timeout at "Perform safe upgrade" stage

[eloquence]

Example:
https://circleci.com/gh/freedomofpress/securedrop/35421#build-timing

Related:
#4740

Output excerpt:

    TASK [common : Check whether tor will be upgraded.] ****************************
    ok: [app-staging]
    ok: [mon-staging]
    
    TASK [common : Perform safe upgrade to ensure all the packages are updated.] ***
    changed: [mon-staging]
make: *** [ci-go] Terminated
Warning: Permanently added '35.247.62.205' (ECDSA) to the list of known hosts.

Too long with no output (exceeded 35m0s)

[emkll]

Also related to #4743

[conorsch]

We have several CI issues reporting "Too long with no output" (see yet another report in #5224 (comment)). First let's get the new CI image in #5360 merged. While I don't expect that to address the root cause of whatever's causing output to break here, the newer package versions may stabilize the runs somewhat.

Regarding the "no output" timeouts, would allocating a tty help? We could add-tt to the call at

securedrop/devops/gce-nested/gce-runner.sh

Line 17 in ebc193c

SSH_OPTS=(-i "$SSH_PRIVKEY" -o "StrictHostKeyChecking=no" -o "UserKnownHostsFile=/dev/null")

That should prevent the ssh output from buffering, which potentially is one reason why we see the timeouts. Why it only happens sometimes, though, I can't say.