move scalability jobs to private pull-through cache and re-enable private clusters

[BenTheElder]

What should be cleaned up or changed:

we should be:

• using the private pull-through cache for giving scale jobs ludicrous quota that we'd ask external users to similarly use a pull-through cache versus hammering registry.k8s.io
• not using registry-sandbox for pause ??
• re-enable private clusters once these are BOTH updated (the pull through is ONLY setup for registry.k8s.io)

Provide any links for context:

#34699 (comment)

/sig k8s-infra testing scalability

... also, we shouldn't be using registry-sandbox ... I should tighten the rate limits there, someone else is going to get the same idea 🙃