RAG: Create document processing pipeline (internal/rag package)

[sre-helmcode]

Description

Create internal/rag/ package with:

• qdrant.go — Qdrant REST client (EnsureCollection, UpsertPoints, Search, DeleteByDocID)
• embedder.go — Ollama embedding client (Embed, EmbedBatch via POST /api/embed, model: nomic-embed-text)
• chunker.go — Recursive character text splitting (1000 chars, 200 overlap)
• parser.go — Document parsing (PDF via ledongthuc/pdf, txt/md, CSV, Excel via excelize, JSON)
• processor.go — Orchestrator (parse → chunk → embed → upsert)

Acceptance Criteria

• All parsers handle their formats correctly
• Chunker produces correct chunks with metadata
• Embedder calls Ollama API correctly
• Qdrant client manages collections and points
• Processor orchestrates the full pipeline
• Unit tests with mocked Ollama/Qdrant