Improved creation and configuration

[Wumpf]

• settings struct that can be changed after the fact
• Overhauled how GpuProfiler is created
  • creation with tracy context is now a separate method which is optional even if tracy feature is active
• test for nesting of profiling scopes with and without support for encoder/pass timer queries
• various docs improvements on the way

The later made it clear to me that there's a bit of love needed to make GpuProfiler useful when using it with multiple command encoder, see #46
After that one is out of the way it's time to get ready for the next wgpu version which allows getting timers directly on compute/render pass boundaries (which can make quite the difference under the hood especially on Metal). This will affect the Scope structs but will otherwise fit in neatly I believe as it's mostly an expansion of what's there, not a replacement :)

[Wumpf]

fyi @cwfitzgerald: this changes the entry point for the tracy integration you added a lil bit

[Wumpf]

@Imberflur: disabling the profiler is back, handled way better than before I think :)

[Imberflur]

Thanks for adding the interface to enabled/disable scopes at runtime ❤️

The overall approach looks pretty good to me. I left a few notes/questions.