Refactor core architecture for extensibility, add SVG→PPTX, Chart→Code features and dockerfile

[kafkayu]

As the title suggests, this update has the following features compared to the current version:

• Refactored core code functions, divided into folders such as autofigure, examples, and docker.

• Added a Dockerfile with instructions on using it.

• To better facilitate Docker deployment, the SAM3 code and required resources have been included.

• To simplify image editing, an SVG to PPTX conversion function has been added, as well as a function to convert local images (without invoking the text-to-image (T2I) module) to SVG to PPT. Example code can be found in AutoFigure-Edit/examples/testchart_local.sh.

• To comprehensively cover paper figures, a chart to Python code conversion function has been added. Users can optionally enable SAM3 as an auxiliary segmentation module, which may improve performance in certain cases.

• Added chart to Python code evaluation, requiring the specification of the reference code path.

Sample Output:
sh examples/testchart_local.sh

============================================================
Paper Method 到 SVG 图标替换流程

Provider: local
输出目录: outputs/chart_demo_nosam
生图模型:
SVG模型: kimi-k2.5

============================================================
步骤一：使用 LLM 生成学术风格图片

Provider: local
模型:
发送请求到: xxx
图片已保存: outputs/chart_demo_nosam/figure.png

============================================================
学术图代码复现模式（不使用 SAM3）：仅根据原图生成 Python 画图代码

============================================================
步骤四（学术图）：多模态调用生成 Python 画图代码

Provider: local
模型: kimi-k2.5
发送多模态请求到: xxx
[Kimi] 图片 Base64 大小: 770728 字符
[Kimi] 发送请求: model=kimi-k2.5, max_tokens=50000
[Kimi] 消息内容包含: 1 个文本, 1 张图片
[Kimi] API 响应状态: completion=True, choices=1
[Kimi] 返回内容长度: 7031
[Kimi] finish_reason: stop
学术图 Python 代码已保存: outputs/chart_demo_nosam/chart_code.py

执行学术图代码脚本: chart_code.py (run_name=initial)
chart_code.py 运行完成，日志: outputs/chart_demo_nosam/chart_code_run_initial.log

执行结果:
figure_path: outputs/chart_demo_nosam/figure.png
chart_code_path: outputs/chart_demo_nosam/chart_code.py
reconstructed_chart_path: outputs/chart_demo_nosam/reconstructed_chart.png

============================================================
步骤七：评估生成的 Python 代码

Provider: local
模型: kimi-k2.5
生成代码: outputs/chart_demo_nosam/chart_code.py
参考代码: /app/examples/inputs/test.py
注意: 图片对比已禁用，仅基于代码进行评估
发送评估请求到: xxx

评分结果已保存: outputs/chart_demo_nosam/evaluation_scores.json

============================================================
评分结果：

参数准确度: 20/25
评价: 数据值基本正确，但(c)(d)(e)的分数有舍入误差(如2.0vs2.05)，ylim设置(115vs110)与参考不符。

视觉相似度: 18/25
评价: 配色接近但存在偏差(如Reference的灰色)，标题位置居中而非左对齐，误差线数值不一致。

代码可执行性: 25/25
评价: 代码结构完整，导入语句齐全，无语法错误，可独立运行并正确保存图像。

总分: 63/75
总体评价: 代码功能完整可运行，数据大体准确，但在视觉细节（颜色精确值、标题对齐方式、数据精度）上与参考代码存在可优化的差异。