v0.4.4

Summary

• Adds high-level raw request handling with validate_arxiv_request(input) and load_arxiv_request(input, main_file=None, refresh=False).
• Accepts bare arXiv IDs, arXiv URLs, Markdown links, and prose while preserving candidate provenance.
• Stops ambiguous requests such as [math/0307200](https://arxiv.org/abs/2609.01574) before any download or active graph replacement.
• Updates CLI and onboarding docs so first-use agents prefer the high-level request tools.

Verification

• python -m pytest -q -p no:cacheprovider -> 294 passed, 1 skipped
• papergraph-mcp --version -> papergraph-mcp 0.4.4
• papergraph-mcp validate-arxiv-request "[math/0307200](https://arxiv.org/abs/2609.01574)" -> status: conflict, action: ask_user_to_choose
• papergraph-mcp load-arxiv-request "[math/0307200](https://arxiv.org/abs/2609.01574)" -> exits 1 with the same conflict JSON