docs(bench): add MacBook Pro M4 Max benchmark results

[gmontana]

Summary

Adds ds4-bench sweep results for a MacBook Pro M4 Max with 128 GB RAM.

The PR adds the raw CSV, generated SVG, and links the new graph from the README.

Benchmark

Hardware reported by Metal:

ds4: Metal device Apple M4 Max, 128.00 GiB RAM

Command:

./ds4-bench \
  -m gguf/DeepSeek-V4-Flash-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8-chat-v2.gguf \
  --prompt-file speed-bench/promessi_sposi.txt \
  --ctx-start 2048 \
  --ctx-max 65536 \
  --step-incr 2048 \
  --gen-tokens 128 \
  --csv speed-bench/macbook_m4_max.csv

Headline results:

• Peak prefill: 160.58 t/s at ctx=6144
• Peak generation: 15.94 t/s at ctx=4096
• 65536 ctx: 109.65 prefill t/s, 11.57 generation t/s
• Average over ctx 49152..65536: 104.72 prefill t/s, 12.53 generation t/s

Validation

• make ds4-bench
• python3 speed-bench/plot_speed.py speed-bench/macbook_m4_max.csv --title "MacBook Pro M4 Max t/s"
• git diff --check