Add PubMed literature mining API

[dainiubi438-debug]

What changed

• add a reusable PubMedClient built on the official NCBI E-utilities API
• expose GET /api/pubmed with validated query and result-limit parameters
• return citation metadata plus publication-year, journal, and publication-type analytics
• support optional PUBMED_EMAIL and NCBI_API_KEY configuration
• document usage and add isolated client and Flask route tests

Why

OpenTCM currently answers from its traditional medicine knowledge graph but has no maintained path for retrieving modern biomedical evidence. This integration provides a small, dependency-neutral backend module that downstream UI and RAG workflows can reuse without coupling PubMed access to the existing LLM implementation.

User and developer impact

Users can retrieve relevant PubMed citations and a lightweight bibliometric summary through a JSON endpoint. Developers can inject a session for testing, configure NCBI identification through environment variables, and handle query errors separately from upstream service failures.

Validation

• python -m unittest discover -s tests -v — 8 tests passed
• python -m compileall -q pubmed_mining.py app.py
• live smoke query: acupuncture osteoarthritis, 2 records returned from 1,456 matches

Operational note

Deployments should set PUBMED_EMAIL and follow the NCBI E-utilities rate limits. NCBI_API_KEY is optional and raises the default request allowance.