conic.h: avoid spurious re-hit (and absorption) on the surface just reflected from

[willend]

Stacked on #2793 (base branch conic-followup-2787). Merge #2793 first, then retarget this PR to main. McStas only.

Problem

After a reflection, a neutron sits on the surface it just left, so that surface's quadric has a root at t = 0. Round-off in solveQuad turned this into small positive roots 1e-9 to 1e-4 m away. traceSingleNeutron then "hit" the same surface again from the outside, and reflectNeutronConic absorbed the neutron (vn > 0).

The bias depended strongly on the figure error. With exact normals about half of the doubly-reflected neutrons were lost, which made results sensitive to a 1 µrad figure error for no physical reason.

Fix

• For the surface hit in the previous step, the next collision now comes from the exact factorisation (C = 0, other root t = −B/A), in new functions getTimeOfNextCollisionConic/Flat().
• traceNeutronConic/Flat() now receive the collision time from traceSingleNeutron instead of recomputing it. Both functions are only called from traceSingleNeutron.
• For all other surfaces nothing changes: the passed t is the value that used to be recomputed.

A post-reflection "nudge" was evaluated and rejected. The spurious roots extend to about 0.1 mm, so a nudge needs about 1 mm to remove all of them, and at that size it already loses genuine hits.

⚠️ Numerical behaviour change

Results of Conics_* change; intensities go up.

Test_Conics_pairs (1e6 rays, seed 1), reflections at the focus (2 m monitor):

Case	Spurious absorptions before → after	Double reflections before → after
OPTIC=1, σ = 0	12 809 → 0	13 198 → 26 007
OPTIC=1, σm = 1 µrad, σs = 0	3 043 → 0	22 947 → 25 990
OPTIC=1, default σ	1 408 → 0	24 582 → 25 990
OPTIC=2 (PH), default σ	3 139 → 0	9 788 → 12 927
OPTIC=4 (PP), default σ	66 → 0	25 942 → 26 008

Reference values (%Example) change by +6% (EH/HE), +33% (PH), +0.2% (PP), and 0% for the fs=8 fi=6 cases. They are updated in this PR.

Tested

• macOS arm64, clang, CPU, 1e6 rays, --seed 1, single process, McStas 3.9.2 with this branch's files.
• Bit-identical to the base branch:
  • Test_NMO and Test_FlatEllipse_finite_mirror, both with default settings and with a silicon substrate (mirror_width=0.0005, mb=0 and mb=4). The FlatSurf part is preventive; no FlatSurf configuration tested showed the problem.
  • Test_Mirror_conic and Test_Mirror_FlatEllipse.
  • WOFSANS width=0.15 (PSD_10 = 3.50e9 ± 0.57e9, stored reference 4e9).
• -Wall -Wextra: no new warnings from conic.h.
• Not tested: Linux, Windows/MSVC, GPU/OpenACC, MPI with more than one process.

🤖 Generated with Claude Code