Adjust release notes layout

Author: mafiesto4

manual/release-notes/1_11/index.md

@@ -26,8 +26,8 @@ This feature comes with a set of great tools, such as Editor's Profiler integrat
 
 ### GPU Particles Performance Improvements
 
-<video src="media/flax-game-particles.mp4" autoplay muted loop></video>
-<em>Game made with Flax running on Nintendo Switch.</em>
+<video src="media/flax-game-particles.mp4" autoplay muted loop width="100%"></video>
+| Game made with Flax running on Nintendo Switch
 
 If your project utilizes GPU particles then on 1.11 it should run a lot of faster! Both rendering and simulation have been refactored to scale better across different hardware. GPU particles sortings run parallel and has very fast-path for small emitters (less than `2048` particles). Finally, GPU execution has been improved by using **UAV overlaps and more optimized memory barriers**.
 
@@ -50,7 +50,7 @@ For this release, we added integration of GPU profiling with [Tracy](../../edito
 ### Visject Improvements
 
 <video src="media/node-formatting-tools.mp4" autoplay muted loop height="400"></video>
-<br><em>New node formatting options</em>
+| New node formatting options
 
 Flax has its own graph editing tools, called Visject. It is used in many areas of the engine, from materials and particles to animations. In 1.11, a lot of new features were added and a lot of already existing ones were improved, making assets that use Visject **even quicker and easy to edit**.
 
@@ -86,8 +86,8 @@ The layers matrix could become quite hard to read when there are lots of layers
 
 ### Threading Improvements
 
-<video src="media/falling-boxes.mp4" autoplay muted loop></video>
-<em>10,000 falling boxes</em>
+<video src="media/falling-boxes.mp4" autoplay muted loop width="100%"></video>
+| 10,000 falling boxes
 
 We've improved many engine parts to use multi-threading more often. Especially, physics simulation results processing, content streaming processing, particles and much more. In the sample falling boxes project, the lowest FPS ancounted when all `10,000` boxes collide went up from average `47fps` to `75fps`. This shows how **engine scales better in large projects** were object count grows exponentionally to meet visual fidelity.