feat(metrics): add model_name label and new throughput/cache metrics

[sufubao]

Summary

Ported metrics improvements from the qwen35 branch to main.

Changes

lightllm/server/metrics/metrics.py

• 所有 Prometheus 指标（Histogram、Counter、Gauge）统一增加 model_name label，支持多模型部署时区分各模型的监控数据
• 新增 counter_inc_by(name, amount) 方法，支持按任意数量递增计数器
• 新增 5 个监控指标：
  • lightllm_prompt_tokens_total — 累计 prefill token 数
  • lightllm_generation_tokens_total — 累计 generation token 数
  • lightllm_cache_hit_rate — 前缀缓存命中率
  • lightllm_gen_throughput — 生成吞吐量（tokens/s）
  • lightllm_num_running_reqs — 当前运行中的请求数

lightllm/server/metrics/manager.py

• MetricServer 新增 exposed_counter_inc_by RPC 方法
• MetricClient 新增 counter_inc_by 异步调用方法

Test plan

• 启动服务后访问 /metrics 端点，确认新指标出现且带有 model_name label
• 发送请求后确认 lightllm_prompt_tokens_total / lightllm_generation_tokens_total 正确递增
• 验证旧有指标（如 lightllm_request_duration）的 label 兼容性

[gemini-code-assist]

Code Review

This pull request introduces several new metrics (such as prompt/generation token totals, cache hit rate, throughput, and running requests) and updates the metrics manager to support incrementing counters by a specific amount. Additionally, all metrics are updated to include a model_name label. The review feedback highlights a critical issue where args.model_name could be None or missing, which would cause runtime errors in prometheus_client when updating metrics. It is recommended to use a safe fallback value like 'unknown' to prevent crashes.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

If args.model_name is None (e.g., if the --model_name argument is not provided at startup) or if the attribute does not exist on args, self.model_name will be None or raise an AttributeError. In prometheus_client, passing None as a label value (e.g., model_name=None) will raise a ValueError: Invalid label value: None at runtime when any metric is updated, crashing the metric server or the background metric thread. To prevent this, use getattr with a safe fallback string like 'unknown'.

Suggested change

	self.model_name = args.model_name
	self.model_name = getattr(args, "model_name", None) or "unknown"