Add a new plugin for command line GPU selection

[SaschaWillems]

Description

This PR adds a new plugin that adds GPU selection via command line. With this plugin, a user can now override the default GPU selection and explicitly request a GPU by a given index. This is esp. useful if you want to debug/test samples on GPUs other than your main GPU. I currently use this to test and fix things using swiftshader as a second ICD on my system. This PR also removes the get_first_gpu instance function that wasn't used anywhere.

The plugin itself is pretty simple, and GPU selection is done as follows:

sample instancing --gpu 1

General Checklist:

Please ensure the following points are checked:

• [x] My code follows the coding style
• [x] I have reviewed file licenses
• [x] I have commented any added functions (in line with Doxygen)
• [x] I have commented any code that could be hard to understand
• [x] My changes do not add any new compiler warnings
• [x] My changes do not add any new validation layer errors or warnings
• [x] I have used existing framework/helper functions where possible
• [x] My changes do not add any regressions
• [x] I have tested every sample to ensure everything runs correctly

• [x] This PR describes the scope and expected impact of the changes I am making

  Note: The Samples CI runs a number of checks including:

  • [x] I have updated the header Copyright to reflect the current year (CI build will fail if Copyright is out of date)
  • [x] My changes build on Windows, Linux, macOS and Android. Otherwise I have documented any exceptions

[asuessenbach]

This plugin seems to be pretty much straight forward... but I'm a bit concerned about the direct access to the [HPP]Instance. Wouldn't it be possible, to carry that information to the sample via the ApplicationOptions?

[SaschaWillems]

Tbh I've had a hard time find out the optimal way of passing plugin data to other classes and just took the same approach as an already existing plugin.

[asuessenbach]

Instead of directly coupling vkb::Instance and vkb::core::HPPInstance to this plugin, you could as well modify Platform::start_app like this:

    auto gpu_selection_it = std::find_if(active_plugins.begin(), active_plugins.end(), [](Plugin *plugin) { return plugin->get_name() == "GPU selection"; });

    std::optional<uint32_t> gpu_selection;
    if (gpu_selection_it != active_plugins.end())
    {
        gpu_selection = reinterpret_cast<::plugins::GpuSelection *>(*gpu_selection_it)->get_gpu_index();
    }

    if (!active_app->prepare({false, window.get(), gpu_selection}))
    {

extending the ApplicationOptions by some std::optional<uint32_t> gpu_selection. And vkb::Instance::get_suitable_gpu (and vkb::core::HPPInstance::get_suitable_gpu) could then take an additional argument std::optional<uint32_t> const &gpu_selection

[SaschaWillems]

Yeah, that would be a different way. But the way plugins pass data and how that data is accessed is all over the place, and every plugin seems to do this differently.

I'd prefer to have this merged (as I need it to fix some open issues) and then at some point talk about how to rework/improve the whole framework situation, esp. for those that work on it.

[tomadamatkinson]

@SaschaWillems we can look at simplifying plugins and allowing them to be enabled pragmatically as well as the CLI. I image we can drop quite a bit of the complexity there