Add frame outcome result to WindowRenderer::render

`WindowRenderer::render` currently returns `()`, so callers cannot tell whether a frame actually reached the window.

For retained or damage-driven apps, that makes it hard to commit paint state correctly: the draw callback may run, but the backend can still fail to acquire, submit, swap, or present the frame afterward.

Would AnyRender be open to a backend-neutral frame outcome API?

```rust
pub enum FrameOutcome {
Presented,
NotReady,
TargetUnavailable,
SurfaceLost,
SurfaceOutdated,
Timeout,
RenderFailed,
SubmitFailed,
PresentFailed,
}
```
The goal is not app-specific dirty tracking. It is exposing renderer/window lifecycle truth so hosts can decide whether to retry.

This would help across backends:

- Vello/WebGPU: surface acquisition, render, submit, present failures
- Skia: surface/context/flush/swap failures
- pixels/softbuffer-style backends: present or target-unavailable states
- Blitz/content renderers: can rely on AnyRender for lifecycle instead of building a parallel model

Desired semantics:

- If no target/surface is available, return without calling the draw callback.
- If drawing runs but the frame does not present, report the failure.
- Presented means the backend believes the frame reached the window/surface.
- Existing fire-and-forget callers could ignore the result.

Start by tracing WindowRenderer::render through the Vello/WebGPU, Skia, pixels/softbuffer-style, and Blitz/content renderer backends mentioned in the issue. Define how each backend maps acquisition, drawing, submission, swapping, and presentation failures to FrameOutcome, while preserving fire-and-forget callers. Done means unavailable targets skip drawing, failed presentation is reported, and successful presentation returns Presented.

Written by the indexing model from the issue text.