#6886 ENG-11157: wait out "being scaled" instead of failing the deploy

Follow-up to #6884. Both "being scaled" refusals during reflex deploy are short-lived — usually the caller's own previous action still rolling out, and the instance-bounds call itself re-rolls a running app, so the submit right behind it meets the same scaling marker. The CLI now retries each at 15s intervals (~2–3 minutes total) with a progress message before giving up, instead of failing a deploy the user will immediately rerun.

Observed live against staging: reflex deploy --min-instances 2 failed with set instance bounds failed: the app is being scaled; change instance bounds after it finishes whenever the previous roll-out hadn't settled.

Linear: ENG-11157

🤖 Generated with Claude Code