Allow initialising shaders in parallel

[DJMcNab]

Headline stats: Reduces shader initialisation time from ~5seconds on Pixel 6 to ~2 seconds (with no compromises on runtime performance)

This is disabled on wasm because there are no threads.
This is disabled on macOS, following the prior art in Bevy (bevyengine/bevy#10812 (comment))

This work queuing system is not intended to be the final form of this, which is why I have made this optional.
This would not be suitable for many consumers, who will likely already have their own thread pool.
We currently just create short lived threads using std::thread::scope.

I still think this is useful to have for experimental purposes, and I think it should also be useful in Xilem in its current form. @xorgy might need you to confirm that

This is not async shader initialisation, which I do also plan to add. This is a useful first step towards that, however

[xorgy]

Is there any reason somebody would want not to initialize in parallel if it is supported on the platform? Memory?

(talking about aside from cases where people have their own thread pooling)

[xorgy]

I still think this is useful to have for experimental purposes, and I think it should also be useful in Xilem in its current form. @xorgy might need you to confirm that

I'll give it a go and let you know.

[DJMcNab]

Is there any reason somebody would want not to initialize in parallel if it is supported on the platform? Memory?

(talking about aside from cases where people have their own thread pooling)

There is also a possibility of starving other threads of the resources they need. And yeah, there is some additional (temporary) memory usage required, although I think (but haven't checked exact numbers) that the amount should be minimal

[waywardmonkeys]

This is disabled on macOS, following the prior art in

I think you left off something...

[DJMcNab]

Thanks, resolved

[ratmice]

Given that the bits I was worried about commitment to semver for it turns out are not actually pub at the package level, this seems fine to me.