Adapt WGPU limits to ggez Backend configuration

[TimerErTim]

Problem

I want to be able to write and use compute shaders in ggez projects, as long as the configured backend allows their usage. Currently this is impossible, because the device limits are always set to downlevel_webgl2_defaults which disallows compute shaders.

Solution

This PR dynamically changes these device limits based on the configured backends. It does this in a way, that is consistent across all possible backends according to configuration. (Backend:All disallows compute shaders because it may target WebGL2, Backend::OnlyPrimary on the other hand always enables them).

[a1phyr]

As I understand wgpu::Limits, not requesting compute shaders in the device will prevent using them even if they are actually supported, is that right ?

[TimerErTim]

Yes, that's right. It limits the wgpu device to the requested limits so that you can't accidentally use the compute shader feature on devices that may not support it.

I used the same philosophy in this PR.

[a1phyr]

An alternative strategy would be to test compute shaders availability in new_from_instance from Adapter::limits, and enable them when requesting a device.

This is more complex but more robust.

[TimerErTim]

An alternative strategy would be to test compute shaders availability in new_from_instance from Adapter::limits, and enable them when requesting a device.

This is more complex but more robust.

That approach would be dynamic, right? With that I mean dependent on the current hardware environment and its GPU capabilities.

If so, I'd avoid that, because any feature (which compute shaders are) usable has to be guaranteed to work under all possible environments, in order to align with wgpu's own philosophy. So either the configuration (in this case Backend::<whatever>) can guarantee that under all circumstances the specific feature (compute shaders) is available, in which case that feature is activated, or it can't guarantee it, in which case the feature is disabled.