fix: Add automatic ECR token refresh mechanism

[aleck31]

Fix ECR token expiration issue by implementing automatic token refresh for Nomad cluster nodes that require ECR access.

Affected nodes and ECR token requirements:

• API节点 (API Node): E2B API服务 ✅ ECR token required
• Client节点 (Client Node): 沙箱执行环境 ✅ ECR token required
• Build节点 (Build Node): 模板构建环境 ✅ ECR token required
• Server节点 (Server Node): Nomad/Consul管理 ❌ No ECR token needed

Modified startup scripts:

1. infra-iac/terraform/scripts/start-api.sh
2. infra-iac/terraform/scripts/start-client.sh
3. infra-iac/terraform/scripts/start-build-cluster.sh

Implementation details:

• Initial ECR token setup during node startup
• Automatic token refresh script (/usr/local/bin/refresh-ecr-token.sh)
• Cron job for periodic refresh (every 10 hours)
• Nomad service restart after token refresh
• Comprehensive error handling and logging
• ECR token validity: 12 hours, refresh interval: 10 hours

This ensures continuous Docker image pulling capability from ECR without manual intervention or service disruption.

Issue #, if available:

Description of changes:

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[jayden-jia]

I'm very happy to receive your PR. This token issue has indeed been a problem that has troubled us for a long time. In commit: 6de8f9e, this issue has been resolved by combining amazon-ecr-credential-helper with Nomad. However, due to time constraints, I only made changes to the API cluster. If possible, you could apply these changes to the other clusters as well.

[aleck31]

This patch includes fixes for the ECR token expiration issues affecting API nodes, client nodes, and build nodes. You can remove the fixed API node-related code while keeping the other modifications.