Benchmark compares speculative decoding methods in llama.cpp: up to 6x speedup

Tests on Qwen 3.6 27B with RTX 6000 PRO show n-gram stack achieves ~6x speedup, DFlash ~3.7x, and MTP ~2.7x on real coding tasks. DFlash support recently merged into llama.cpp (PR #22105).