Turn on fillet_gn_section_generator

solidcore's one new UpstreamFixes flag since c9a06dc7c (#2260, #2261,
#2262; WR-685, solidcore#2250): the curvature section is the Gⁿ
generator's, at the equal-area thumbweight by default, and a variable
round is the union of its balls, a law whose balls fold refused as
NoRoomReason::BallsFold.

Toggled on the rig on solidcore (flag off against on, everything else as
pinned): only curvature and variable rounds move. With it off the
curvature rounds fail or come out wrong (the hole rim's r1 curvature
round 63.94 mm³ against 15207.56, the boss root's 420.9 against
26056.15) and the variable cone case takes the great circles
(11955.6466); with it on all match their closed forms (next two
commits). Every other case keeps its status and its volume bit for bit.
Rig time 1064 s off, 1248 s on, the difference the curvature rounds that
now build (the hole rim 89 s and boss root 92 s, as on SMLib d01d7b0's
102 s and 105 s). One unit test, three_rounds_at_a_corner, moves: the
corner's variable transition between rounds of 2, 3, 4 meets its
neighbours at 0.144° (the test allows 0.05°), as on SMLib d01d7b0, its
volume and patch corners as before; listed, as for SMLib. Not broken by
the flag rule: HELD_BACK stays empty.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SdcgEu4Lr3moCCisdZShKz

crates/smlib/src/core.rs

@@ -365,6 +365,7 @@ pub(crate) fn fixes() -> UpstreamFixes {
     f.point_class_gap_3d_propagates_no_drop_error = true;
     f.fillet_surface_plane_curve_on_plane_equation = true;
     f.fillet_extended_rail_end_at_side_edge_crossing = true;
+    f.fillet_gn_section_generator = true;
     f
 }