feat: add Semantic Scholar citation traversal and author tools

[RaggedR]

Summary

• Adds 6 new MCP tools for Semantic Scholar integration: get_paper_details, get_citations, get_references, find_related, search_author, get_author_papers
• Enables forward and reverse citation traversal — essential for literature reviews
• No changes to existing ArXiv tools or dependencies (httpx is already in pyproject.toml)

New Tools

Tool	S2 Endpoint	Purpose
get_paper_details	/paper/{id}	Full metadata: title, authors, citations, TL;DR, abstract
get_citations	/paper/{id}/citations	Forward traversal — who cites this paper
get_references	/paper/{id}/references	Reverse traversal — what this paper cites
find_related	/recommendations/v1/papers/forpaper/{id}	Embedding-based similar paper discovery
search_author	/author/search	Find authors by name with h-index, affiliations
get_author_papers	/author/{id}/papers	List an author's publications

All tools accept flexible paper identifiers: S2 hash, ArXiv:XXXX.XXXXX, DOI:..., or CorpusID:....

Configuration

Optional SEMANTIC_SCHOLAR_API_KEY environment variable for higher rate limits (5,000 vs 100 req/5min). Works without it.

Implementation

• New file: src/arxiv_mcp/tools/semantic_scholar.py (self-contained, ~300 lines)
• Modified: src/arxiv_mcp/main.py (import + register 6 tools)
• Uses httpx.AsyncClient (already a dependency) for async S2 API calls
• Follows the project's existing pattern: async functions returning formatted strings
• Error handling via try/except with logger, matching existing tool patterns

Test plan

• All imports resolve correctly
• No changes to existing tools or dependencies
• Follows existing code patterns (async functions, logger, markdown output)

🤖 Generated with Claude Code