feat: Add ArXiv channel for academic paper search (zero-config)

Summary

Add ArXiv as a zero-config channel for academic paper search. ArXiv is the primary preprint server for AI/ML, CS, physics, and math — and its API requires no authentication.

Why

Agent-Reach is heavily used for research workflows (competitive analysis, trend discovery, literature surveys). Adding ArXiv closes a gap for anyone researching academic topics, tracking new AI papers, or doing literature reviews. Currently, users must manually search ArXiv via web — structured API access would be a major unlock.

Proposed Implementation

Use the official ArXiv API:

• Search papers: http://export.arxiv.org/api/query (no auth)
• Paper details: http://export.arxiv.org/api/query?id_list=XXXX.XXXXX (no auth)
• Category browse: cat:cs.AI, cat:cs.CL, cat:stat.ML, etc. (no auth)
• Author search: au:Author_Name (no auth)

Tier: 0 (zero-config, zero-auth) — works immediately after install.

Acceptance Criteria

• agent-reach doctor shows ArXiv as ok out of the box
• Channel supports: full-text search, paper detail (title, abstract, authors, categories, PDF link), date-range filtering, category filtering
• Results return structured JSON with: title, authors, abstract, categories, published date, PDF URL, ArXiv ID
• SKILL.md documents the new channel with curl examples
• Tests cover: API mocking, URL matching, health check states

Additional Context

• ArXiv API docs: https://arxiv.org/help/api
• Categories list: https://arxiv.org/category_taxonomy
• Rate limit: 1 request/3s (self-enforced, documented)
• Estimated user value: high — essential for AI/ML research, academic trend tracking