feat: enrich guidance with tool recommendations and strict html-path output

feat: refactor usage-insights to qualitative review workflow

feat: prioritize long conversations in session sampling

feat: add concrete session evidence to Karpathy evaluation

fix: percentage display and add evidence-based evaluation requirements

refactor: simplify Karpathy section and reorder report structure

feat: prioritize AI expert assessment, add detailed scoring rubric

refactor: remove keyword-based scoring, switch to AI expert analysis mode

docs: 更新 README.md，添加 Karpathy 推文链接和完整文档

feat: 添加 Karpathy Agentic Engineering 评分系统

feat: 添加抽样复盘功能，抽取最近20个session详细内容

refactor: 移除外部 API key 依赖，改为证据数据 + Agent 分析模式

feat: initial public release of usage-insights hybrid analyzer