修复 MMKG 子图/QA 三元组解析并补全示例图片

[W-RMSL]

No description provided.

[Copilot]

Pull request overview

This PR fixes parsing/mapping logic for MMKG visual triples when generating subgraph-based QA pairs, and ensures the multimodal pipeline example assets (including images) are included in source distributions.

Changes:

• Include all files under dataflow/example/ in the package manifest (and fix the example path comment).
• Fix MMKG visual triple parsing to correctly extract image IDs from the <obj> ... <rel> segment and preserve first-seen order for URL pairing.
• Update the MultimodalKGPipeline example input to contain a non-empty img_dict and corresponding vis_url entries.

Reviewed changes

Copilot reviewed 4 out of 6 changed files in this pull request and generated no comments.

File	Description
MANIFEST.in	Ensures example files (incl. images) are shipped in the sdist by recursively including dataflow/example/*.
dataflow/operators/multi_model_kg/generate/mmkg_visual_triple_subgraph_qa_generator.py	Fixes image-id extraction from vis_triple and aligns de-dup ordering with vis_url for correct img_dict construction.
dataflow/operators/multi_model_kg/filter/mmkg_visual_triple_subgraph_sampling.py	Corrects regex boundaries to parse <subj> up to <obj> and <obj> up to <rel> for visual triples.
dataflow/example/MultimodalKGPipeline/input.json	Adds concrete example image mappings and URLs so the example is runnable with bundled assets.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.