Split prepare_windows span to communicate when we're waiting for vsync

> prepare_windows usually means it's either waiting on the / bottlenecked by the GPU, or just waiting for vsync
> So if GPU work is sub 1ms, then you're probably just hitting the display's refresh wait

By @JMS55.

This is an important and non-obvious quirk of our tracy profiles.

Ideally we could have `waiting_for_vsync` and `waiting_for_gpu` spans here to make the problem (or lack thereof!) much more obvious to users getting started with profiling their apps.

Start by locating the prepare_windows span in the rendering or frame-pacing code and inspect how it distinguishes GPU work from the display refresh wait. Define separate waiting_for_vsync and waiting_for_gpu spans so Tracy profiles communicate which wait is occurring, then verify the resulting profiling behavior.

Written by the indexing model from the issue text.