feat: add router architecture deployment configs for multi-model production setup

[blefo]

New AI model services and orchestration:

• Added qwen3_coder_30b_gpu service to docker-compose.nilai-router-1.yml for deploying the Qwen3-Coder-30B-A3B-Instruct model with GPU support and health checks.
• Added gpt_oss_20b_gpu and qwen3_thinking_4b_gpu services to docker-compose.nilai-router-2.yml for deploying the GPT-OSS-20B and Qwen3-4B-Thinking-2507 models, including inter-service health dependencies.
• Added arch_router_1_5b_gpu and qwen3_vl_4b_gpu services to docker-compose.nilai-router-3.yml for deploying the Arch-Router-1.5B and Qwen3-VL-4B-Instruct models, with multimodal support and health dependencies.