Fix all failing CI on main: concurrency smoke, MEX parity, example_dock shading, wiki links

[HanSur94]

What

All three workflows that were red on main are fixed. Diagnosed by a fan-out of agents (one per failing workflow, each adversarially verified), then every fix was re-verified locally in MATLAB R2025b.

Workflow (was failing)	Root cause	Fix
Tests — Concurrency Smoke (ubuntu, macOS)	install() only builds MEX inside the needs_build() gate, which probes the committed binary_search_mex. lockfile_mex is the one MEX not committed, so a clean checkout never compiled it and TestConcurrencyIntegration failed.	install.m: best-effort ensure-block that compiles lockfile_mex independent of the gate (honours FASTSENSE_SKIP_BUILD, no-ops when present, try/catch → pure-MATLAB sidecar fallback).
Tests — Octave parity + TestMexParity	The minmax tail-anchor (commit 31d04b7f) made minmax_core emit 2*nb+1 points; the parity reference oracles still expected 2*nb.	Update both oracles (tests/test_mex_parity.m + tests/suite/TestMexParity.m) to mirror the shipped MEX/fallback. The class-suite copy was a latent failure on macOS/Windows MATLAB.
Example Smoke — example_dock	Same tail-anchor: the two shaded-region boundaries downsample to different lengths when one is a constant baseline (addFill), so patch() got mismatched XData/YData ("Vectors must be the same length").	libs/FastSense/FastSense.m: each boundary keeps its own downsampled X (xd/xd2); shared cache trimmed to a common length in render and updateShadings.
Wiki Link Check	5 Wiki-Browser pages used bare extensionless links [Event Viewer](Event-Viewer); lychee resolves relative links as file paths and 404s (16 errors).	Append .md to every bare target. In-app WikiBrowser strips trailing .md, so navigation is unaffected.

Note: the Tests/parity and Example-Smoke failures are two consumers of the same regression (31d04b7f's tail-anchor).

Verification (MATLAB R2025b, this branch)

• test_mex_parity → PASS (incl. minmax_core_mex parity, the exact assertion that was red)
• TestMexParity → 9/9 pass (incl. testMinmaxCoreParity)
• example_dock → renders all 5 tabs / ~21M pts, incl. the formerly-failing "Power Systems" tile
• install() → runs clean; guard confirmed reachable on a clean checkout
• mh_style / mh_lint / mh_metric --ci → clean on all changed files; checkcode adds no new issues
• Wiki: all 24 link targets now resolve to existing .md files; #refresh-paths anchor left intact

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[HanSur94]

Patch coverage addressed in b4f953f. Added two regression tests to TestAddShaded that execute all 15 previously-uncovered lines: render()'s cache branch (n>20000) and mid-size branch (5000<n≤20000), plus updateShadings() via a zoom. A monotonic upper curve over a constant fill baseline makes the two boundaries downsample to different lengths (the minmax tail-anchor), so each test would have thrown "Vectors must be the same length." on the pre-fix code. Codecov re-evaluates on the new commit.

ℹ️ The MATLAB Tests (Q-Z) red on the prior commit was TestTagPerfRegression/testSensorTagGetxyGate — a relative-timing perf gate (overhead_pct <= 5) in SensorThreshold, untouched by this PR and historically flaky on shared CI runners. The fresh run should clear it.