Track Responsive-Design carousel ResizeObserver event counts

[issackjohn]

Follow-up from WebKit#453 review.

Review context:

• Reviewer comment: github dot com/Experimental Workload: Responsive Design WebKit/Speedometer#453#discussion_r3189824107
• My follow-up promise: github dot com/Experimental Workload: Responsive Design WebKit/Speedometer#453#discussion_r3205139859

Current state:

• experimental/responsive-design/src/lib/components/recipe-carousel.js uses ResizeObserver to update _carouselWidth, which affects _getItemsPerView() and carousel scroll distance.
• Other Responsive-Design components also intentionally use browser-observable responsive APIs:
  • app-ribbon.js uses ResizeObserver for visible button count.
  • information-window.js uses ResizeObserver.
  • recipe-grid.js uses matchMedia("(max-width: 640px)") and emits resize-work-complete for benchmark synchronization.
• The benchmark resize steps wait for one resize-work-complete event from recipe-grid, but they do not verify how many ResizeObserver callbacks fire in recipe-carousel or whether the count is consistent across browsers.

Problem to investigate:

• ResizeObserver callback counts can differ by browser or layout timing.
• If recipe-carousel sees a different number of resize notifications, the workload may do inconsistent responsive work even when the benchmark still completes.

Suggested direction:

• Track the number of ResizeObserver notifications in recipe-carousel during ReduceWidthIn5Steps and IncreaseWidthIn5Steps.
• Decide whether this should be:
  • a diagnostic console.warn when counts differ from expected values, or
  • a strict unittest/e2e assertion if the expected count is stable across Safari, Chrome, Firefox, and Edge.
• If strictness is too brittle, expose a debug counter and validate only coarse invariants, e.g. callbacks occurred and final _carouselWidth matches the final iframe width class/threshold.
• Keep benchmark measurement boundaries unchanged unless the count tracking shows missing async work.

Validation:

• Run Responsive-Design across the browser matrix and compare observed ResizeObserver counts.
• Add a unit/e2e test only after expected browser behavior is confirmed stable enough.