docs(benchmarks): K8V4 result + P2P-CNS finding on lucebox-hub dual-GPU (#229)

K8V4 same-card bench (-ctk q8_0 -ctv q4_0): 74.68 mean tok/s vs 73.97
default KV — basically identical (+1%). KV-format optimization doesn't
help at HumanEval-scale (<150-tok prompts × 128-tok gen) where the KV
pool isn't the bottleneck. Asymmetric quant available via PR #56/#54
(merged 2026-04-28 in lucebox-hub).

P2P-CNS finding (more important): the dual-GPU split bench from
cb089e1 ran on a chipset that reports "Chipset Not Supported" for
GPU↔GPU peer access (PHB topology, common consumer-board limitation).
The lucebox-hub split path requires CUDA P2P for direct draft-feature
transfers; without it, falls back to host-staging copies (CPU↔GPU
bouncing). The +1.7% we observed is therefore NOT a fair test of the
split's value — it's measuring same-card vs same-card-with-host-staging
overhead.

@weicj's published 51.86 tok/s on dual 2080 Ti 22GB (PR #80) presumably
ran with P2P available. Our negative result on PHB-only consumer boards
is rig-specific, not a refutation of the technique. Updated BENCHMARKS
row to reflect this honestly.

Setup gotcha now documented in BENCHMARKS: check `nvidia-smi topo -p2p r`
before configuring --target-gpu / --draft-gpu. NVLink-bonded setups
would typically expose P2P (cross-rig confirmation needed on lucebox
specifically; @JusefPol's NVLink win was on vLLM TP=2, not lucebox).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

results/lucebox-k8v4-20260504-144647/phase-c-k8v4.log

@@ -0,0 +1,24 @@
+[bench] target = /mnt/models/gguf/qwen3.5-27b/Qwen3.5-27B-Q4_K_M.gguf
+[bench] draft  = /mnt/models/huggingface/qwen3.5-27b-dflash/model.safetensors
+[bench] bin    = /opt/ai/lucebox-hub/dflash/build/test_dflash
+[bench] tmp    = /tmp/dflash_bench
+[bench] skipping tokenize (reusing /tmp/dflash_bench)
+
+[bench] mode=fast  n_gen=128
+prompt                         steps     AL   pct%    tok/s
+--------------------------------------------------------------
+  has_close_elements              28   4.57   28.8    55.54
+  separate_paren_groups           20   6.40   40.9    75.72
+  truncate_number                 22   5.82   38.1    70.09
+  below_zero                      18   7.11   45.1    82.68
+  mean_absolute_deviation         24   5.33   33.3    63.61
+  intersperse                     28   4.57   29.2    54.61
+  parse_nested_parens             13   9.85   63.5   109.09
+  filter_by_substring             21   6.10   38.4    72.09
+  sum_product                     16   8.00   55.5    91.51
+  rolling_max                     21   6.10   41.4    71.83
+--------------------------------------------------------------
+MEAN                                   6.38   41.4    74.68
+
+commit/step range: 4.57 - 9.85
+tok/s range:        54.6 - 109.1