feat: IGLA RAG Engine Cycle 15 — Local Retrieval Augmented Generation

- Document indexing with file type detection (Zig, Python, JS, Markdown, Text)
- Auto-chunking with chunk type detection (Function, Struct, Comment, etc.)
- TF-IDF based similarity search for relevance scoring
- Top-K context retrieval with source attribution
- Wraps CodeSandboxEngine (Cycle 14) for layered architecture

Benchmark Results:
- Documents indexed: 3
- Chunks indexed: 6
- Speed: 6947 ops/s
- Improvement rate: 1.00 (PASSED > 0.618)
- Tests: 182/182 passing

phi^2 + 1/phi^2 = 3 = TRINITY | KOSCHEI RETRIEVES ALL | CYCLE 15

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: gHashTag