Add arXiv paper ingestion via API

[dentity007]

What needs to be done

Create forge/ingest/arxiv_loader.py that fetches papers from arXiv by author name or search query using the arXiv API, extracts the abstract and full text (from PDF), and ingests into the knowledge base.

Why this matters

arXiv is the primary source of AI/ML research papers. Researchers need to ingest papers directly.

Where to look

• forge/ingest/document_loader.py (existing pattern for PDF loading)
• forge/ingest/upserter.py (batch embedding + upsert)
• examples/ingest_arxiv.py (example script already exists as a template)

NVIDIA Stack Impact

Uses Triton for batch embedding of paper chunks.

Acceptance criteria

• Can ingest papers by author name
• Can ingest papers by search query
• PDF text extraction works
• Chunks embedded and upserted to Qdrant
• Graph nodes created for papers and authors
• Tests added