Add many_meshlet_materials stress test example

[CrazyRoka]

Objective

• Introduce a dedicated example to measure performance overhead when dealing with a high number of meshlet instances and unique materials.

Solution

• Added a new example many_meshlet_materials that spawns a configurable grid of meshlet bunnies.

Testing

• SystemInfo { os: "Linux (CachyOS Linux rolling)", kernel: "6.19.10-1-cachyos", cpu: "AMD Ryzen 7 5800H with Radeon Graphics", core_count: "8", memory: "15.0 GiB" }
• AdapterInfo { name: "AMD Radeon Graphics (RADV RENOIR)", vendor: 4098, device: 5688, device_type: IntegratedGpu, device_pci_bus_id: "0000:03:00.0", driver: "radv", driver_info: "Mesa 26.0.3-arch2.2", backend: Vulkan, subgroup_min_size: 64, subgroup_max_size: 64, transient_saves_memory: false }
• cargo run --features=meshlet,https --release --example many_meshlet_materials

---

Showcase

Example works perfectly with 5x5 grid.

The issue begins when grid count increases to 10x10. Meshes start flickering.

With 50x50 meshes, the whole screen flickers.

When running with unique materials, performance significantly drops on my hardware.

[JMS55]

The flickering seems like a regression - I've tested with a much greater number of instances before.

Assuming it's not CPU bound, there's a lot of simple things we can do to improve the perf of many materials. Currently we spawn a full screen triangle for every single material, which is super inefficient. We should be spawning a quad covering only the needed pixels per material. I just haven't gotten around to it.