fix(skill-build): use --host/--port for thv serve and correct health-check path

[ChrisJBurns]

Summary

Fixes the skill-build action, which is broken at runtime with Error: unknown flag: --address. Verified against the ToolHive v0.29.1 source (cmd/thv/app/server.go, pkg/api/server.go, pkg/api/v1/healthcheck.go).

Closes #16.

Root cause

The action installs thv@latest and starts a daemon to serve thv skill build/thv skill push. It passed CI at merge time (2026-03-10) but upstream thv later drifted, breaking three things — only the first is fatal:

1. thv serve --address no longer exists 🔴 — serve registers --host (default 127.0.0.1) and --port (default 8080), not --address. This is the reported hard failure.
2. Wrong health-check path 🟠 — the readiness probe polled /api/v1beta/health, but the health route is mounted at top-level /health. The old path 404s and would have timed out the 30s wait even after fixing the flag.
3. Health gated on container runtime 🟡 — /health returns 503 unless a container runtime (Docker/Podman) is running, and curl -sf fails on 503. Skill build/push only package OCI artifacts (via go-containerregistry) and don't need a runtime, so readiness should not depend on it.

Changes

• thv serve --address "127.0.0.1:${PORT}" → thv serve --host "127.0.0.1" --port "${PORT}"
• Readiness probes now hit /health and accept any HTTP response (including 503) as "daemon up" — which also unblocks runners without Docker (relates to macOS: Container runtime not available for Toolhive #3).
• Added a daily schedule trigger to test.yml so future breakage from upstream thv@latest releases is caught proactively rather than at the next push. (The test-skill-build-action job already existed; it simply hadn't re-run since the upstream change.)

Testing

• thv serve flags, the /health mount point, and the 503-without-runtime behavior were confirmed directly in the ToolHive v0.29.1 source.
• CI on this PR exercises test-skill-build-action end-to-end against thv@latest.

🤖 Generated with Claude Code