fix(ui): latitude-inverse denominator + reticle city-hover gate (PR #187 findings)

Hoshe's blocking find: district_pos_at divided the latitude axis by
tex_h where the server forward map uses ta.h.saturating_sub(1)
(district_profile.rs:1436; ground-truth inverse aliveness_probe.rs:511)
— every off-equator click descended into the wrong district, up to
~20km drift, live-repro'd at 10.2km. Fixed to /(tex_h - 1.0) with the
reference implementation's degenerate-texture guard; verified against
the repro (row 50: old formula recovered 41, fixed recovers 50 exact).
Columns were already correct (longitude wraps — the asymmetry the
docstring documented but the code didn't implement).

Regression tests transcribe the server forward formula (source cited)
and round-trip three off-equator rows — one per hemisphere plus
near-pole — asserting EXACT recovery; a dedicated drift guard is
framed as Tyre's C2 (a client-side twin of a server mapping owns its
own round-trip proof). Root-caused the coverage hole: the old
equator test sampled tex_h*0.5 believing it was the equator pixel —
the true equator is (tex_h-1)*0.5, so the test never verified what it
claimed; corrected with the why documented.

Araminta's find: the descend reticle drew unconditionally on hover —
over a city the marker flared white (city-click signal) while the
reticle+extent label promised descent, though the click correctly
resolved to city data. Extracted _should_draw_descend_reticle() with
the missing _hovered_city.is_empty() clause; three state tests pin
shows-over-map / hides-over-city / hides-when-idle.

test_atlas_descend_entry 32/32; adjacent clusters no ripple; gdlint
clean (atlas_viewer.gd at exactly the 1000-line cap).

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

client/ui/implant/apps/atlas/atlas_descend_geometry.gd

@@ -80,17 +80,30 @@ static func reticle_label(center: Vector2) -> Dictionary:
 ## function, per the amendment's §5 carry-over wording. The server's forward
 ## mapping (body has a radius) is:
 ##   px = ((dx * DISTRICT_M) / circumference_m mod 1.0) * tex_w
-##   py = (0.5 + clamp(dy * DISTRICT_M / meridian_m, -0.5, 0.5)) * tex_h
-## which is a linear scaling of the same world-metre fraction the equatorial/
-## meridian district COUNT already IS (district_cols = round(circumference_m
-## / DISTRICT_M), the same value build_district_grid()'s `cols` converges to
-## for a body tiled edge-to-edge) — so inverting is "pixel fraction * district
-## count", not a re-derivation of the server's geodesy. Self-contained: does
-## NOT depend on district_grid (the whole-body layer) having arrived yet, so
-## descent works immediately on entry even before that async layer resolves.
-## The "no radius" branch (tiny test bodies, body_radius_km absent/<=0)
-## mirrors the server's own fallback: the district grid IS the heightmap
-## grid 1:1.
+##   py = (0.5 + clamp(dy * DISTRICT_M / meridian_m, -0.5, 0.5)) * ta.h.saturating_sub(1)
+## (district_profile.rs:1436 — note `.saturating_sub(1)`, NOT a bare `ta.h`;
+## confirmed against the ground-truth inverse aliveness_probe.rs:511 too:
+## `row / (ta_h - 1) - 0.5`). Columns and rows are DELIBERATELY asymmetric:
+## longitude WRAPS (rem_euclid), so a column has no "last pixel" edge case and
+## divides by the plain width; latitude CLAMPS at the poles, so row 0 and row
+## (h-1) are real, distinct endpoints (the north/south pole pixels) and the
+## division must land exactly on them — dividing by `tex_h` instead of
+## `tex_h - 1` introduces a systematic drift that grows with |lat_frac|
+## (worst at the poles, invisible only at the exact equator row, `tex_h*0.5`,
+## where the two formulas round identically). This is why inverting is "pixel
+## fraction * district count" for columns but NOT a symmetric operation for
+## rows — the row inverse must undo the SAME -1 the forward map applied.
+##
+## The (wx/circumference_m, wy/meridian_m) fractions are linear scalings of
+## the same world-metre quantity the equatorial/meridian district COUNT
+## already IS (district_cols = round(circumference_m / DISTRICT_M), the same
+## value build_district_grid()'s `cols` converges to for a body tiled
+## edge-to-edge) — so inverting is "pixel fraction * district count", not a
+## re-derivation of the server's geodesy. Self-contained: does NOT depend on
+## district_grid (the whole-body layer) having arrived yet, so descent works
+## immediately on entry even before that async layer resolves. The "no
+## radius" branch (tiny test bodies, body_radius_km absent/<=0) mirrors the
+## server's own fallback: the district grid IS the heightmap grid 1:1.
 static func district_pos_at(
 	canvas_pt: Vector2, tex_w: float, tex_h: float, body_radius_km: float
 ) -> Vector2i:
@@ -103,5 +116,12 @@ static func district_pos_at(
 	var district_cols: float = roundf(circumference_m / DISTRICT_M)
 	var district_rows_half: float = roundf((meridian_m / DISTRICT_M) * 0.5)
 	var col: int = roundi((canvas_pt.x / tex_w) * district_cols)
-	var row: int = roundi(((canvas_pt.y / tex_h) - 0.5) * district_rows_half * 2.0)
+	# tex_h - 1.0, matching the forward map's ta.h.saturating_sub(1) — NOT a
+	# bare tex_h (see the docstring above; this was a live bug, T-1138 PR #187
+	# review, Hoshe: every off-equator click descended into the wrong district).
+	# Guarded the same way the server's own inverse (aliveness_probe.rs:510,
+	# `if ta_h > 1 { ... } else { 0.0 }`) guards the same division — a
+	# degenerate 1px-tall texture would otherwise divide by zero.
+	var lat_frac: float = ((canvas_pt.y / (tex_h - 1.0)) - 0.5) if tex_h > 1.0 else 0.0
+	var row: int = roundi(lat_frac * district_rows_half * 2.0)
 	return Vector2i(col, row)

client/ui/implant/apps/atlas/atlas_viewer.gd

@@ -16,8 +16,7 @@ extends Control
 ##   - Empty markers case (server #832/#833 not yet shipped): bare heightmap renders
 ##     fine, no sidebar opens, overlays draw nothing.
 ##   - Overlays (#836) plug into _overlay_visibility dict and _draw_overlays().
-##   - T-1138: view is FIXED (set_view() is capture-API-only); descent screen has the reticle math.
-## Navigation: Hover reticle · Click descend · Click city data (wins) · R reset(no-op) · Esc back
+## Navigation (view FIXED, T-1138): hover reticle (hidden/city) · click descend/city · esc back
 
 signal back_pressed
 signal economics_link_requested(system_id: String)
@@ -645,13 +644,15 @@ func screen_to_canvas(screen_point: Vector2) -> Vector2:
 
 func _draw() -> void:
 	draw_rect(Rect2(Vector2.ZERO, get_rect().size), COLOR_BG)
-	if _hover_active and _heightmap_texture != null:
+	if _should_draw_descend_reticle():
 		_draw_descend_reticle()
 
 
-## T-1138 NOT-TO-SCALE reticle (§5's open design point). Draw calls only —
-## geometry/rationale in atlas_descend_geometry.gd (draw_line must run on `self`).
-func _draw_descend_reticle() -> void:
+func _should_draw_descend_reticle() -> bool:  # hidden over a city — _try_click_city wins the click
+	return _hover_active and _heightmap_texture != null and _hovered_city.is_empty()
+
+
+func _draw_descend_reticle() -> void:  # draw calls only — geometry in atlas_descend_geometry.gd
 	var center: Vector2 = _hover_screen_pos
 	for seg: Array in AtlasDescendGeometry.reticle_segments(center):
 		draw_line(seg[0], seg[1], AtlasDescendGeometry.COLOR_DESCEND_RETICLE, 1.5)