docs(bench): add a Current status table, realign three stale verdicts

[kalwalt]

Why

bench/README.md had grown into a 677-line chronological log: six sections, each with its own status box. Answering "where do things stand today?" meant reading all six and assembling the answer — and every new measurement had to be threaded into six places, which is exactly how it drifted out of date.

What

A Current status table at the top, measured over four full-suite runs on merged dev (after #159, #165, #166 — idle machine, warm-up discarded). It states plainly that the sections below are the historical record and their numbers are not updated, so future measurements touch one table instead of six sections.

case	4 runs	status
linalg.lu_solve	1.43 / 1.49 / 1.46 / 1.47	open — tightest signal in the suite
homography2d.check_subset	1.35 / 1.27 / 1.28 / 1.63	open — new matmath()
affine2d.run 3 pts	1.34 / 1.31 / 1.30 / 1.41	open — same cause
math.get_gaussian_kernel size 7	1.23 / 1.21 / 1.16 / 1.15	open — 4/4, at/just above floor
yape06 / yape / invert_3x3 / svd_*	≤ 1.17	resolved
everything else	±1.12	noise

Three stale verdicts corrected

yape06 was recorded as "unchanged, still open" at 1.14–1.47. On merged dev it measures ~1.04 across four runs. Marked resolved — but explicitly unexplained: #166's alias alone did not produce this (that PR measured and reported no benefit for yape06), so something in the combination of the three fixes did. Recorded as unexplained rather than credited to any one change.

get_gaussian_kernel size 7 was described as "mostly below floor", with the next sentence asserting everything else sat at or below ~1.15x. The later series puts it at 1.15–1.23, four of four favouring jsfeat — at or just above the floor, not below it. Both spots corrected; carried as open.

lu_solve's verdict was right but its numbers were superseded: now 1.43–1.49 with a ±0.03 spread, tighter than YAPE ever was. It has never been profiled, and #159 predicted in advance that its fix could not explain it (lu_solve constructs no internal matrix_t).

Scope

Documentation only — no source or bench code touched. npm test 265 passed, license-check 93 files clean.

Note: dist/ and types/ show local modifications from the verification build on this machine; they are deliberately not included here.

Refs #86.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds a single “Current status” summary at the top of bench/README.md so readers can see today’s benchmark findings without re-reading multiple historical sections, and realigns a few stale narrative verdicts to match recent measurements on merged dev.

Changes:

• Adds a “Current status” table summarizing four full-suite runs on merged dev, and clarifies that the rest of the document is historical (not continuously updated).
• Updates the YAPE section framing and the lu_solve / get_gaussian_kernel size 7 notes to reflect newer measurements and revised verdicts.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.