Gate OSWEC refinement against published motion

[cmccomb]

What changed

Keep a compact, unrounded fixture from the untouched published OSWEC nonlinear visualization run. Before using the output-disabled refinement run to assess step sensitivity, require its 0.1 s source output to match every published flap pose, velocity, acceleration, and principal force sample exactly.

Evidence

• Compared run 38259283403 (matlab-oswec-nonlinear-viz-pressure) with refinement run 38263477931 (dt-0.1.mat): all 1,201 samples of time and seven six-channel histories matched exactly.
• Local fixture replay passed the new array identity assertions. git diff --check and Python compilation passed.
• This strengthens the evidence chain for the existing refined-step motion gate. The published 0.1 s independent Python motion remains qualified; badge counts do not change.