feat: add graceful-shutdown-implementer skill and zero-downtime-deployment-expert agent

[ArthurAdrianoMM]

What

Adds the graceful-shutdown-implementer skill and the zero-downtime-deployment-expert agent.

• Skill: implements SIGTERM/SIGINT handling with the correct drain sequence (stop intake, finish in-flight work, close resources, exit), fixes containers that never receive signals (shell-form CMD, missing init), and aligns app drain timeouts with stop_grace_period / terminationGracePeriodSeconds. Node and Python examples included.
• Agent: a zero-downtime deployment persona covering rolling updates, blue-green, canary releases, connection draining, and N/N+1 compatibility - refusing to design rollouts until health checks and graceful shutdown prerequisites exist.

Why

Dropped requests during deploys almost always trace back to missing graceful shutdown - a prerequisite no existing skill covers, and the natural companion to deployment strategy guidance.

Checklist

• npm run skill:validate passes for the new skill (note: an unrelated pre-existing skill aws-cloudwatch-investigation currently fails validation on staged)
• npm run build executed to update the README
• No duplicate of existing skills/agents