Generate the kahlus-bench "detection-power sweep" figure — a clean multi-panel matplotlib comparison of several methods across a swept parameter, with a black dashed analytic-truth reference line, per-method colored marker lines, panel titles that carry direction ("lower is better"), and a one-sentence suptitle that states the takeaway. Use when the user wants a plot that looks like the kahlus-bench power-sweep / Independent-benchmark figure, wants to compare estimators/models across a parameter grid (coupling strength, sample size, noise, etc.), or says "make a figure like the kahlus-bench sweep", "that power sweep plot", or references that figure's look for any method-v…
