fix: reduce ALB target group deregistration delay from 600s to 30s

[jesserockz]

Summary

The ALB target group deregistration_delay is set to the AWS default of 600 seconds (10 minutes). This means every rolling deployment waits up to 10 minutes for the old task to drain before completing, even though the service handles short-lived webhook requests with no long-running connections.

Change

.modules/webservice/network.tf: deregistration_delay 600 → 30 seconds

Reasoning

The service-hub-bots workload is entirely stateless — inbound traffic is GitHub webhooks and CLA sign callbacks, all of which complete well within a few seconds. A 30 second drain window is more than sufficient and matches the actual connection lifecycle.

Impact

• Deployments complete ~10 minutes faster
• No cost saving directly, but reduces the window where two Fargate tasks run simultaneously during a rolling update (deployment_maximum_percent = 200), which marginally reduces Fargate compute usage per deploy
• No risk to in-flight requests — any active connection will complete within the 30 second window