CI: Set retry to 0 for publish jobs to avoid spot instance interruptions

Summary

This MR sets retry: 0 for all CI jobs in the "publish" stage to prevent automatic retries when spot instances are interrupted. This ensures that publish operations (which should be idempotent and manually retriggered if needed) don't run multiple times due to infrastructure failures.

Background

This MR is a follow-up to !18661 (merged) which introduced default retry policies for spot instance failures. During review, it was identified that publish jobs require different retry behavior and should be extracted to a separate, focused MR for clarity.

Changes

Publish Jobs Modified:

• apt_repo_debian
• apt_repo_ubuntu
• docker:merge_manifests
• docker:promote_to_latest
• documentation:publish
• gitlab:octez-evm-node-release
• gitlab:publish
• gitlab:release
• grafazos.gitlab:release
• opam:release
• publish:release-page
• rollup-node.gitlab:octez-smart-rollup-node-release
• rpm_repo_fedora
• rpm_repo_rockylinux
• teztale.gitlab:release
• teztale.publish:release-page

Each job now explicitly sets:

~retry:Gitlab_ci.Types.{max = 0; when_ = []}