Fix headless_renderer example and mention Screenshot.

[kpreid]

Objective

• Makes headless_renderer example work instead of exiting without effect.
• Guides users who actually just need Screenshot to use that instead.

This PR was inspired by my own efforts to do headless rendering, in which the complexity of the headless_renderer example was a distraction, and this comment from #12478 (comment) :

The example added in #13006 would benefit from this change: be sure to clean it up when tackling this work :)

That “cleanup” was not done, and I thought to do it, but it seems to me that using Screenshot (in its current form) in the example would not be correct, because — if I understand correctly — the example is trying to, potentially, capture many consecutive frames, whereas Screenshot by itself gives no means to capture multiple frames without gaps or duplicates. But perhaps I am wrong (the code is complex and not clearly documented), or perhaps that feature isn’t worth preserving. In that case, let me know and I will revise this PR.

Solution

• Added exit_condition: bevy::window::ExitCondition::DontExit
• Added a link to Screenshot in the crate documentation.

Testing

• Ran the example and confirmed that it now writes an image file and then exits.

[alice-i-cecile]

But perhaps I am wrong (the code is complex and not clearly documented), or perhaps that feature isn’t worth preserving. In that case, let me know and I will revise this PR.

The feature is definitely worth preserving: that's important for machine learning applications, and some forms of testing.

I'm not familiar enough to know if we could reuse Screenshot to do this perfectly and more simply, but this is a clear improvement over the current state of things. As a result, merging :)

[IceSentry]

what makes it exit then?

[kpreid]

That's very much worth explaining, so I added a comment about that. The code explicitly generates an AppExit event.

I didn’t see any information in the contribution guide about force-push/rebase/squash preferences, so I added an extra commit with that comment, but I would personally rather rebase to 2 commits. Please let me know what is preferred.

[alice-i-cecile]

We don't have strong preferences about git workflows (primarily to reduce the barrier of entry), but I personally find that rebasing / force pushing sucks on Github because it risks stranding comments and generating nonsense notifications.