Which post-stall model should POLY cd hand over to? The blend dips from 0.183 to 0.145 over 20–24° #388

With the blend from #386, a POLY (Breukels) section's cd falls from 0.183 at 20° to a minimum of 0.145 at 24.3° before it rises along the flat plate. So dcd/dα is negative over 20°–24°. The cause is that the two models disagree at 20°: lei_poly_coeffs(0.1, 0.08) gives cd 0.183 from the polynomial against 0.080 from 2|sin α|³, and cl 1.07 against 0.22. Any blend between them has to fall. The 5° width only sets how fast.

The question is which post-stall model the Breukels polar should hand over to. Options:

A) Keep the flat-plate forms 2 sin α |sin α| cos α and 2 |sin α|³ (what #386 does, and what the Python VSM does), and accept the cd dip.
B) Anchor the post-stall cd to the polynomial's 20° value, for example cd(20°) + (2|sin α|³ − 2 sin³(20°)). That makes cd monotone, but it shifts every post-stall drag by about +0.10.
C) Use a different post-stall model (Viterna or similar) matched to cl and cd at 20°.

My recommendation is A until a measured LEI post-stall polar is available to choose between B and C. This needs your call on which model is intended. #386 does not change the model.