[feat] add port conflict detection for cache-queue-port and engine-wor…

[sunlei1024]

Motivation

为提升系统在多组件并行启动场景下的稳定性，本 PR 新增了 cache-queue-port 与 engine-worker-queue-port 的端口冲突检测逻辑。当用户显式指定端口且端口已被占用时，系统将提前识别并快速失败及提示，避免在启动hang住，从而提升整体用户体验与可用性。

Modifications

• 新增对 cache-queue-port 的端口可用性检查逻辑
• 新增对 engine-worker-queue-port 的端口可用性检查逻辑
• 优化错误提示信息，明确冲突来源与解决方式
• 增强 Engine 相关模块在多实例部署时的健壮性

Command

当前服务启动时，可指定四类端口

python -m fastdeploy.entrypoints.openai.api_server \
       --model /workspace/ERNIE-4.5-0.3B-Paddle \
       --port 8680 \
       --engine-worker-queue-port 8386 \
       --cache-queue-port 8683 \
       --metrics-port 8682 \

端口冲突示例

Traceback (most recent call last):
  File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/workspace/FastDeploy/fastdeploy/entrypoints/openai/api_server.py", line 738, in <module>
    main()
  File "/workspace/FastDeploy/fastdeploy/entrypoints/openai/api_server.py", line 717, in main
    if not load_engine():
  File "/workspace/FastDeploy/fastdeploy/entrypoints/openai/api_server.py", line 123, in load_engine
    engine = LLMEngine.from_engine_args(engine_args)
  File "/workspace/FastDeploy/fastdeploy/engine/engine.py", line 75, in from_engine_args
    config = engine_args.create_engine_config()
  File "/workspace/FastDeploy/fastdeploy/engine/args_utils.py", line 1229, in create_engine_config
    assert is_port_available(
AssertionError: The parameter `engine_worker_queue_port`:['8386'] is already in use.

Accuracy Tests

该改动不涉及模型推理逻辑或 kernel 修改，对模型精度无影响。

Checklist

• Add at least a tag in the PR title.
• Code formatted; pre-commit 已通过
• Add unit tests
• No accuracy-related changes

[paddle-bot]

Thanks for your contribution!

[gongshaotian]

需要处理下冲突