Order-balanced performance proof for Moppe canopy and cloud lighting

Follow-up performance proof for #GL2MX9 used a detached build of the untouched 19fa6f9 checkpoint and two adjacent 32-configuration A/B pairs in opposite order, with 15 settle and 60 measured frames per configuration. Averaging each order before comparing produced a median configuration delta of -0.134 ms (-0.97%), with 19 of 32 configurations faster; the fully enabled delta was -0.046 ms. Treat these small differences as performance parity, not a speedup. The order effect by itself was much larger than the implementation delta, confirming that the earlier non-adjacent 2.76% result was too sensitive to run order. 5BNBQQ X9PHPN