Disable PA on/off ramps; restore mainline to pre-ramp state

Three iterations of the Daikoku PA ramps never read as a real
interchange. Measured why, rather than tuning further:

  * the entry and exit decks INTERSECTED -- 2547 sample pairs within
    30 studs horizontally at <12 studs dY, worst case 0.5 studs apart
    at dY=0.0. On screen that is the bowtie tangle of white rails.
  * both real Daikoku ramps turn the SAME way (clockwise; the guide
    video says "right, right"). This build mirrored them into each
    other.
  * the real arc wraps OUTSIDE the lot and forms its north/east
    boundary. This build cut across the apron interior.

Gated behind BUILD_PA_RAMPS=false rather than deleted -- the radius
was already correct (470 studs vs an OSM-measured mean of 469) and
the merge lands tangent at 0.1-0.2 deg divergence. Both worth keeping.

Also removed two leftovers that would have outlived the ramps:

  * rampOpen punched holes in the mainline's outer parapet, railing
    and kill wall for ramps that no longer exist -- a barrier gap with
    a drop behind it. Sealed via RAMP_EDGE_OPENINGS=false. Verified:
    Barrier/EdgeRail/KillWall = 1646 each = 823 samples x 2 edges,
    0 road samples without a guard within 90 studs.
  * gore zebra hatching painted 1024 bars + 64 boundary lines marking
    a junction that is not there. Sealed via RAMP_GORE_MARKINGS=false.

Verified after a cache-busting fresh rebuild (Studio's require() cache
otherwise serves the previous map and reports a false pass):
  parts 22628 -> 15988
  all ramp part classes = 0
  Road = 823 (unchanged), spawn pad + PASpawn intact
  clip probe TRUSTWORTHY (control +11/-0), CLIPPING=0

Research and rebuild spec kept for when this is revisited:
reference/ramps/BUILD_SPEC.md, INDEX.md (10 CC-licensed Commons
photos), docs/daikoku-reference-spec.md (OSM circle fits).

tools/probe_floating.luau

@@ -0,0 +1,139 @@
+-- FLOATING SCENERY PROBE.
+--
+-- The wide captures show grey slabs/towers hanging in the sky with nothing
+-- under them. No existing probe catches this: the clip probe only asks
+-- "does road intersect building", which is 0 and correct. Floating geometry
+-- is the opposite failure -- parts that touch NOTHING at all.
+--
+-- Metric: for each candidate scenery part, raycast straight DOWN from its
+-- underside and report the distance to the first thing hit. A part sitting
+-- on ground/deck returns ~0. A floating part returns a large number or nil
+-- (nothing beneath it at all).
+--
+-- CONTROL: the same ray cast from a part known to be seated (a Road part)
+-- must return a small distance. If the control comes back nil/huge the
+-- raycast itself is broken and every "floating" result is meaningless --
+-- the silent-zero trap that produced two phantom defects earlier.
+
+local ws = game:GetService("Workspace")
+local map = ws:FindFirstChild("Map")
+if not map then
+	return "NO MAP"
+end
+
+local candidates, roads = {}, {}
+for _, d in ipairs(map:GetDescendants()) do
+	if d:IsA("BasePart") then
+		local nm = d.Name
+		if nm:sub(1, 5) == "Road_" then
+			table.insert(roads, d)
+		elseif
+			nm:find("Building")
+			or nm:find("Tower")
+			or nm:find("Skyline")
+			or nm:find("Block")
+			or nm:find("Slab")
+			or nm:find("Distant")
+		then
+			table.insert(candidates, d)
+		end
+	end
+end
+
+local params = RaycastParams.new()
+params.FilterType = Enum.RaycastFilterType.Exclude
+
+-- distance from a part's underside straight down to the first hit
+local function dropDist(p)
+	params.FilterDescendantsInstances = { p }
+	local origin = p.Position - Vector3.new(0, p.Size.Y / 2 + 1, 0)
+	local res = ws:Raycast(origin, Vector3.new(0, -6000, 0), params)
+	if not res then
+		return nil
+	end
+	return (origin - res.Position).Magnitude
+end
+
+local out = {}
+
+-- CONTROL: a road part must find something close beneath it (deck/pier/ground).
+local ctl, ctlN = nil, "none"
+for _, rd in ipairs(roads) do
+	local d = dropDist(rd)
+	if d then
+		ctl, ctlN = d, rd.Name
+		break
+	end
+end
+if ctl == nil then
+	table.insert(out, "CONTROL=FAIL raycast found nothing under ANY road -- probe is broken, ignore results below")
+else
+	table.insert(out, ("CONTROL=PASS %.1f studs under %s (raycast works)"):format(ctl, ctlN))
+end
+
+-- Classify candidates.
+local FLOAT_MIN = 60 -- studs of empty air below = floating
+local floating, seated, voidBelow = 0, 0, 0
+local worst, worstNm, worstPos = -1, "none", nil
+local samples = {}
+for _, p in ipairs(candidates) do
+	local d = dropDist(p)
+	if d == nil then
+		voidBelow += 1
+		if #samples < 8 then
+			table.insert(
+				samples,
+				("%s@%.0f,%.0f,%.0f NOTHING-BELOW"):format(p.Name, p.Position.X, p.Position.Y, p.Position.Z)
+			)
+		end
+	elseif d > FLOAT_MIN then
+		floating += 1
+		if d > worst then
+			worst, worstNm, worstPos = d, p.Name, p.Position
+		end
+		if #samples < 8 then
+			table.insert(
+				samples,
+				("%s@%.0f,%.0f,%.0f drop=%.0f"):format(p.Name, p.Position.X, p.Position.Y, p.Position.Z, d)
+			)
+		end
+	else
+		seated += 1
+	end
+end
+
+table.insert(
+	out,
+	("candidates=%d roads=%d | seated=%d floating(>%d)=%d nothingBelow=%d"):format(
+		#candidates,
+		#roads,
+		seated,
+		FLOAT_MIN,
+		floating,
+		voidBelow
+	)
+)
+if worstPos then
+	table.insert(
+		out,
+		("WORST %s drop=%.0f at %.0f,%.0f,%.0f"):format(worstNm, worst, worstPos.X, worstPos.Y, worstPos.Z)
+	)
+end
+for _, s in ipairs(samples) do
+	table.insert(out, "  " .. s)
+end
+
+-- Name histogram so we can see WHICH generator made them.
+local hist = {}
+for _, p in ipairs(candidates) do
+	local base = p.Name:gsub("_%d+$", "")
+	hist[base] = (hist[base] or 0) + 1
+end
+local hl = {}
+for k, v in pairs(hist) do
+	table.insert(hl, ("%s=%d"):format(k, v))
+end
+table.sort(hl)
+table.insert(out, "byName: " .. table.concat(hl, " "))
+
+return table.concat(out, "\n")