fix #116 (partial, Ghidra-confirmed): slide_sphere degenerate guard uses F_EPSILON, not EpsilonSq

The user brought up Ghidra; its decompiler (patchmem.gpr, full PDB)
resolved the Binary-Ninja `test ah,5` x87 branch-sign ambiguity that
blocked the desk read. CSphere::slide_sphere (0x00537440) decompiles
cleanly to:

  fVar3 = |cross(collisionNormal, contactPlane.N)|²;
  if (::F_EPSILON <= fVar3) {                       // crease exists
      ... offset = cross * dot(cross,gDelta)/fVar3;
      if (|offset|² < ::F_EPSILON) return COLLIDED_TS;   // degenerate guard
      ... add_offset_to_check_pos -> SLID_TS
  }

Retail compares the SQUARED magnitudes against F_EPSILON
(0.000199999995 ~= 0.0002 = PhysicsGlobals.EPSILON). Our port compared
against EpsilonSq (0.0002^2 = 4e-8) - a ~5000x too-tight threshold (the
BN pseudo-C rendered the comparison as `test ah,5` after an x87 FCMP,
which is sign-ambiguous; agent reads disagreed). Fixed both comparisons
at TransitionTypes.cs:3098,3105 to EPSILON.

Effect: crease-exists now needs >=0.81 deg between the wall and contact
normals (was 0.011 deg - which routed near-parallel pairs through the
numerically unstable projection); the degenerate guard now hard-stops
slides under ~1.41 cm like retail (was 0.2 mm). Branch POLARITY was
already correct - no change there.

No regression: full physics suite (612) + full Core (1443) green. Not a
register deviation (no row existed; this is an undocumented porting
error corrected to match retail).

This does NOT close #116 - it fixes a tangential constant, not either
reported shape. Ghidra also settled the two shapes' diagnosis (recorded
in ISSUES.md #116 + physics digest):
- Shape-1: our cn=UnitZ default IS retail-faithful (validate_transition
  0x0050aa70 has the identical `if (collision_normal_valid==0)
  set_collision_normal(UnitZ)`). The real divergence is upstream -
  tick-22760 our collision_normal_valid was false where retail's was
  true (it recorded the door-face normal). Needs the instrumented
  tick-22760 replay.
- Shape-2 (D4 stays skipped, note sharpened): slide_sphere slides
  in-frame (SLID_TS) so Z=1.92 is faithful and the D4 Z=2.0 hard-stop
  pin is the suspect half; the threshold fix didn't move D4 (real slide,
  not degenerate). Needs a cdb trace of an airborne wall hit.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

docs/ISSUES.md

@@ -3930,13 +3930,47 @@ retail's viewer-distance smoothing (update_viewer region) before touching.
 
 ## #116 — Slide-response divergence family: near-perpendicular lateral slide lost + first-airborne-frame in-frame slide vs hard stop
 
-**Status:** OPEN
+**Status:** OPEN (narrowed) — one Ghidra-confirmed faithfulness fix
+SHIPPED 2026-06-12; both reported shapes still need a runtime trace.
 **Severity:** LOW-MEDIUM (over-blocking, never under-blocking — no
 walk-throughs; feel-level divergence at walls/doors)
 **Filed:** 2026-06-11 (BR-7 / A6.P4 ship session)
 **Component:** physics (slide response — `SlideSphere` degenerate-offset
 guard + first-contact-frame behavior)
 
+**GHIDRA SESSION 2026-06-12 (the BN branch-sign ambiguity RESOLVED via a
+second decompiler — Ghidra MCP, patchmem.gpr, full PDB):**
+- **SHIPPED (faithfulness fix):** `CSphere::slide_sphere` (Ghidra
+  `0x00537440`) compares its SQUARED magnitudes against `::F_EPSILON`
+  (= 0.000199999995 ≈ 0.0002 = `PhysicsGlobals.EPSILON`): `if (::F_EPSILON
+  <= |cross|²)` (crease) and `if (|offset|² < ::F_EPSILON) return
+  COLLIDED_TS` (degenerate guard). Our port compared against `EpsilonSq`
+  (0.0002² = 4e-8) — a ~5000× too-tight threshold (the BN `test ah,5`
+  obscured it). Fixed at `TransitionTypes.cs:3098,3105`; full physics
+  suite (612) + full Core (1443) green, no regression. Crease now needs
+  ≥0.81° between normals (was 0.011°); the guard stops slides under
+  ~1.41 cm like retail (was 0.2 mm). NOT a register deviation (no row
+  existed — it was an undocumented porting error; the fix matches retail).
+  ⚠️ This does NOT fix either reported shape below.
+- **Shape-1 RE-DIAGNOSED — our `cn=UnitZ` default is RETAIL-FAITHFUL.**
+  Ghidra `validate_transition` (`0x0050aa70`) does exactly our
+  `TransitionTypes.cs:3701-3702`: `if (collision_normal_valid == 0)
+  set_collision_normal(UnitZ)`. So the harness `cn=(0,0,1)` is the
+  faithful FALLBACK; the real divergence is UPSTREAM — at tick-22760 our
+  `collision_normal_valid` was FALSE (→ UnitZ) where retail's was TRUE
+  (it had recorded the door-face normal `(0,+1,0)`). The bug is in the
+  COLLISION-RECORDING path (find_collisions / collide_with_environment),
+  not slide/validate. Next: replay tick-22760
+  (`DoorBugTrajectoryReplayTests`) instrumented to see where our
+  collision-normal recording drops the wall normal.
+- **Shape-2 NARROWED — D4 stays skipped.** Ghidra confirms slide_sphere
+  applies the slide IN-FRAME (`add_offset_to_check_pos` → SLID_TS), so our
+  Z=1.92 is faithful TO slide_sphere and the D4 Z=2.0 hard-stop pin is the
+  SUSPECT half. But the threshold fix did NOT change D4 (its offset is a
+  real slide, not degenerate), so whether retail's first airborne frame
+  REACHES slide_sphere (→1.92) or hard-stops upstream still needs a cdb
+  trace of an airborne wall hit before flipping the assertion.
+
 **Two pinned shapes, both pre-dating BR-7 (the per-cell shadow port left
 them byte-identical):**