Add Qwen3.8 Max to live bench

[anthonypizzolato-abacus]

Adds Qwen3.8 Max to live bench runs.

Field	Value
Model	Qwen3.8-Max (Alibaba flagship, released 2026-08-02)
Provider API ID	qwen3.8-max (rolling GA id on DashScope; no dated snapshot published)
Serving endpoint	Alibaba DashScope first-party, OpenAI-compatible: https://dashscope-intl.aliyuncs.com/compatible-mode/v1
Context window	1,000,000 tokens (max output 65,536)
Pricing	$2.00/M input, $6.00/M output (GA flat rate)

Changes (mirrors the deepseek-chat / DeepSeekAPI first-party pattern)

• lcb_runner/lm_styles.py — new LMStyle.QwenAPI; LanguageModel("qwen3.8-max", "Qwen3.8-Max", LMStyle.QwenAPI, datetime(2026, 8, 2), link="https://qwen.ai/blog?id=qwen3.8") grouped at the end of the Qwen-family block
• lcb_runner/runner/qwen_runner.py — new QwenRunner (exact mirror of DeepSeekRunner; env key DASHSCOPE_API, DashScope OpenAI-compatible base URL)
• lcb_runner/runner/runner_utils.py — LMStyle.QwenAPI → QwenRunner
• lcb_runner/prompts/code_generation.py — QwenAPI uses the generic chat prompt (same as OpenAIChat/DeepSeekAPI)

Verified: LanguageModelStore['qwen3.8-max'] resolves and all edited files parse. Runner requires DASHSCOPE_API env var at run time.

Do NOT merge until reviewed.