fix(actions): try to recover if just fails to install

[ledif]

I have noticed an increase in workflow failures due to flaky errors with installing just. For example, in the past few days, the following jobs to build images have failed:

• https://github.com/ublue-os/bluefin/actions/runs/12600726942/job/35120374239
• https://github.com/ublue-os/bluefin/actions/runs/12576191989/job/35052038346
• https://github.com/ublue-os/bluefin/actions/runs/12567338911/job/35033772133

The command curl -L https://api.github.com/repos/casey/just/releases/latest | jq -r '.tag_name' is sometimes returning "null" instead of an actual version, but it's difficult to tell if the curl call is failing or if it's returning a payload that is being parsed incorrectly.

In either case, adding --retry to curl and if it's still null, hardcoding a known good version should do the trick for now.

[tulilirockz]

I believe we should try using that setup-just action at some point, but this is a very nice idea

[m2Giles]

This should be more resistant

while [[ "${JUST_VERSION:-}" =~ null || -z "${JUST_VERSION:-}" ]]
do
    JUST_VERSION=$(curl -L https://api.github.com/repos/casey/just/releases/latest | jq -r '.tag_name')
done