Arxiv ID validation warns but does not reject invalid IDs + HTTP used instead of HTTPS for API

[consigcody94]

Found via code audit. scripts/fetch_paper.py:55-56 normalize_arxiv_id() prints WARNING but returns untouched input for invalid IDs. Line 68 uses http:// not https:// for arxiv API query. Also no requirements.txt for the skill scripts (requests, pymupdf4llm undeclared).