Wgpu-rs: Prevent any drop() logic if inside a panic

[kvark]

It's annoying to see a panic inside a panic if we are seeing drop() failing because we are in process of stack unwinding. We should check for this in drop() implementations and avoid doing anything that can fail.

[yanchith]

I'd like to take this. Will post progress reports and eventual questions in this thread.

[yanchith]

So browsing through the Drop impls, 3 seem they could use this:

• impl<'a> Drop for SwapChainOutput<'a>
• impl<'a> Drop for ComputePass<'a>
• impl<'a> Drop for RenderPass<'a>

They have in common that they act more like RAII guards with more logic than just destructors happening on drop. I personally only ever encountered the SwapChainOutput double panic in the wild.

@kvark do you think we should be preventing the drop logic for all three, or just SwapChainOutput? I also might have missed a place where this would be useful...

EDIT: Just tried inserting a couple a panic!() calls and it seems that preventing the drop logic in SwapChainOutput is enough to prevent the abort.

[kvark]

In all three, if the drop is happening inside a panic, just don't do anything.