Sum a span's cursor by name, and guard a scroll's re-clamp

`cursor` added `px` and `rel` by hand where the placing loop below now says
`fixed += len.without_leftover()` -- the same sum, one of them named. And
`let along = total` shadowed the closure that makes a span along the row,
two meanings for one word in one function; the local said nothing `total`
did not.

A scroll's draw writes `amt` and `snap_end`, so a second draw at another
viewport reads what the first wrote. Warm still matches cold because
re-clamping is idempotent, but nothing said so and nothing checked it: the
seed scans build scrolls and never scroll them. The test scrolls four
distances, one past the end, and widens.

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

tests/cases/unsettled.rs

@@ -1067,3 +1067,32 @@ fn resizing_under_a_short_scroll_snaps_its_window_tall_content_again() {
 
     assert_same_regions(&warm, &ids, &cold, &cold_ids);
 }
+
+/// A scroll clamps its position against the box it is drawn in, so drawing it
+/// once at one viewport and again at another writes state the second draw then
+/// reads. That the answer is still the one a cold layout gives is a property
+/// of the clamp, not something the layout enforces.
+#[test]
+fn a_scrolled_view_resized_lands_where_a_cold_layout_puts_it() {
+    for amt in [10.0, 40.0, 90.0, 140.0] {
+        let mut warm = Harness::new((100, 100));
+        let (_, warm_scroll) = plant_wider(&mut warm, 100.0);
+        warm.move_to((50.0, 50.0));
+        warm.scroll((-amt, 0.0));
+        warm.frame();
+        warm.resize((160, 100));
+        warm.frame();
+
+        let mut cold = Harness::new((160, 100));
+        let (_, cold_scroll) = plant_wider(&mut cold, 100.0);
+        cold.move_to((50.0, 50.0));
+        cold.scroll((-amt, 0.0));
+        cold.frame();
+
+        assert_eq!(
+            warm.region(&warm_scroll),
+            cold.region(&cold_scroll),
+            "scrolled by {amt} then widened"
+        );
+    }
+}