ci: add docker-publish workflow so dispatch registers on main

[arnaubennassar]

Why this is separate from PR #24

PR #24 (feat/aggkit-backend) is large, still in draft, and not ready to merge. However, .github/workflows/docker-publish.yaml needs to exist on the default branch (main) before GitHub will register its workflow_dispatch and release triggers — that registration only happens for workflow files present on the default branch, regardless of what other branches contain.

The exact failure that motivated this

$ gh workflow run docker-publish.yaml --repo agglayer/agglayer-dev-ui --ref feat/aggkit-backend
HTTP 404: workflow docker-publish.yaml not found on the default branch

What this PR does (and doesn't do)

• Adds only .github/workflows/docker-publish.yaml, byte-identical to the copy on feat/aggkit-backend (PR feat: migrate bridge backend from Bridge Hub API to aggkit bridge service (+ bridge tracker UI, rc5) #24, commit dd070d7).
• This is a no-op until someone manually dispatches it or publishes a GitHub release — merging it changes no existing behavior and builds nothing on its own.
• The actual Docker image build/publish logic this workflow performs lives on feat/aggkit-backend (PR feat: migrate bridge backend from Bridge Hub API to aggkit bridge service (+ bridge tracker UI, rc5) #24) and will be reviewed there.
• docker-pr.yaml is intentionally not included here — it's a pull_request-triggered workflow that runs from the PR head, so it doesn't need to be on main.

After this merges

Dispatching against a feature branch (e.g. gh workflow run docker-publish.yaml --ref feat/aggkit-backend) will run that branch's copy of the workflow and check out that branch — the copy on main only needs to exist and be valid YAML to unlock registration; it doesn't need the Dockerfile or other PR #24 files to be present on main.

---

🤖 Generated with Claude Code

https://claude.ai/code/session_01SUseLAAcYiPPkqrUqFehtY

[claude]

Code review

1 issue found. Checked for bugs and CLAUDE.md compliance.

Issue: Inline shell logic in workflow YAML (team standards violation)

File: .github/workflows/docker-publish.yaml, lines 135-197 and 243-280

The team standards require:

Keep workflow YAML thin — put logic in scripts. Any non-trivial shell logic (more than a single command) must live under .github/scripts/ or .github/actions/<name>/ so it can be run and debugged locally. YAML steps are just invocations. Never split logic across many steps sharing state via $GITHUB_OUTPUT.

This workflow has two steps with substantial inline shell:

• "Compute image tags" (lines 135-197): ~43 lines of shell (semver parsing, prerelease detection, tag construction, collision guard)
• "Smoke test pushed image" (lines 243-280): ~38 lines of shell (container run, polling loop, HTTP checks)

Additionally, "Compute image tags" writes to $GITHUB_OUTPUT for cross-step state sharing, which the rule explicitly prohibits.

Suggestion: Extract into .github/scripts/compute-image-tags.sh and .github/scripts/smoke-test.sh.