Add M2 Ultra (192 GB) benchmark

[timtan]

Thanks for the awesome work. here is the benchmark for my M2Ultra.

Hardware: Apple M2 Ultra, 24 cores (16P + 8E), 192 GB unified memory. Metal backend, IQ2XXS w2Q2K Q8 imatrix GGUF.

Ran the command from speed-bench/README.md:

./ds4-bench -m ds4flash.gguf --prompt-file speed-bench/promessi_sposi.txt \
  --ctx-start 2048 --ctx-max 65536 --step-incr 2048 --gen-tokens 128 \
  --csv speed-bench/m2_ultra.csv

Then:

python3 speed-bench/plot_speed.py speed-bench/m2_ultra.csv --title "M2 Ultra t/s"

Range across 32 frontiers: prefill 411 → 270 t/s, gen 23.2 → 20.4 t/s. Prefill consistently higher than M4 Max at every context; gen slightly lower (more GPU cores, lower per-core clock).