Ability to execute shaders from within shaders outside of raytracing

Currently we can (and are required to) execute shaders from within shaders with vulkan raytracing https://nvpro-samples.github.io/vk_raytracing_tutorial_KHR/

And Nvidia devices have supported [this functionality](https://developer.nvidia.com/blog/cuda-dynamic-parallelism-api-principles/) in compute for the better part of a decade.

Metal [also supports this functionality](https://developer.apple.com/videos/play/wwdc2020/10013/) in the kind of general context asked for here.

This feature request asks for the ability to execute, at the very least *compute shaders* from within other compute shaders.

Start by reading the Vulkan ray-tracing tutorial and the linked CUDA Dynamic Parallelism and Metal materials to compare the requested capability. Then inspect the Vulkan-Docs specification and related tools to determine where such a feature would need to be defined. Done means the scope and API requirements for launching compute shaders from compute shaders are specified.

Written by the indexing model from the issue text.