Build on wgpu 30

Two majors, and the renderer is under everything else that is left to
extract, so it goes before the slices that would otherwise be written twice.
`image` 0.25.6 -> 0.25.10 rides along; `winit` stays on 0.30.12, since 0.31
is only a prerelease.

What the API asked for:

- An instance now takes the display it will present on, and GLES on Wayland
  needs it -- so the window the surface is made from is handed over with it.
- `get_current_texture` returns a `CurrentSurfaceTexture` rather than a
  `Result`, which replaces an `unwrap` that would have panicked on a resize
  or an occluded window: reconfigure when the surface is outdated, lost or
  suboptimal, and skip the frame when there is nothing to draw into.
- Presenting moved to the queue, after `pre_present_notify` as before.
- Bind group layouts and vertex buffer layouts are sparse now, so a pipeline
  states `Some(layout)` per slot.

Verified with the three render checks and `tests/draw_cost.rs`: identical
pictures, and 33.6/167/587/2855 us per frame at 8/64/256/1024 layers against
33.3/161/588/2903 on wgpu 28.

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

Cargo.toml

@@ -29,9 +29,9 @@ edition = "2024"
 [workspace.dependencies]
 pollster = "0.4.0"
 winit = "0.30.12"
-wgpu = "28.0.0"
+wgpu = "30.0.1"
 bytemuck = "1.23.1"
-image = "0.25.6"
+image = "0.25.10"
 parley = "0.11.1"
 swash = "0.2.10"
 fxhash = "0.2.1"