Add llm-cpp: 26 single-header C++20 LLM infrastructure libraries

[Mattbusel]

Summary

Adds llm-cpp to the Artificial Intelligence section.

llm-cpp is a collection of 26 single-header C++20 libraries covering the full LLM inference stack:

• Token streaming and quantization (INT4/INT8/FP16)
• KV-cache management and SIMD batch inference
• Speculative decoding and draft-verify pipelines
• Financial time-series with special-relativistic geometry (Lorentz transforms on OHLCV data)

All headers are MIT licensed, C++20, zero external dependencies, header-only drop-in.

🤖 Generated with Claude Code