Hidden WKWebView throttles JS timers, skewing perf probes

When driving a web page in an off-screen, alphaValue=0 NSWindow WKWebView (a headless-style test harness), WebKit throttles setTimeout/setInterval to about 1 Hz. A 50 ms sampler showed ~950 ms delays even on a nearly empty page, and polling loops like while (!cond()) await sleep(100) take about 10x longer than their timeout suggests. Don't read event-loop lag or wall time from timer-based loops in such a harness. Timestamp the real completion event instead (e.g. wrap the page's state-update callback and record performance.now() when the answer arrives), and always run a trivial control case to calibrate. Also double-check done-conditions: an undo after a marker move restores the same rows, so waiting for a row count to drop never finishes and looks like a hang.