Improve performance for many entities

[Wumpf]

What

• Fixes Slowdown with 1'000 entities #2285
  (.. at least fixes it well enough for the moment I reckon)

Improves both the per space view rendering and the space view adding heuristic. In particular on the later there's still significant gains to be made.
Most importantly we should be scaling now much better with number of added systems.

Went fairly early with the decision to have the primary datastructure to pass around be a Map<System, Vec<Entity>> (conceptual) instead of the other way round since this is what we need in the systems most of the time.
This is another step towards a stronger contract of systems specifying what components they will query ahead of time!

The way it is implemented on thr (rename from BlueprintTree 💥) SpaceViewContents also paves the way for ui selection of systems for a given entity path which complements the ongoing work on our new data ingestion interfaces.

---

Testcase:
examples/python/open_photogrammetry_format/main.py --no-frames, fully reset viewer and then hiding NOT REMOVING the points (the world/pcl entity) and deselecting them.
(The bold marked pieces are very important as they have an influence on what heuristics run - see #3077)

Before:

After:

(release runs, averaged over a bunch of frames on my M1 Max)

Highlights:

• AppState::show: 24.0ms ➡️ 14.6ms
  • SpaceViewBlueprint::scene_ui: 4.9ms ➡️ 2.6ms
    • this was the original primary optimization target!
  • Viewport::on_frame_start: 15.2ms ➡️ 7.4ms
    • default_created_space_view: 12.7ms ➡️ 5.1ms
      • quite some time spent on this, but we want to do a paradigm shift there: Move heuristic for viable space view "candidates" to space view class #3079

(slight number discrepancies from the screenshot are due to doing a different run)

---

Draft todo: Code sanity check. Suspecting some heuristics might be slightly broken, need to go through examples

Checklist

• I have read and agree to Contributor Guide and the Code of Conduct
• I've included a screenshot or gif (if applicable)
• I have tested demo.rerun.io (if applicable)

• PR Build Summary
• Docs preview
• Examples preview
• Recent benchmark results
• Wasm size tracking

[emilk]

cargo r -p raw_mesh is also a pretty good benchmark

[github-actions]

Size changes

Name	Previous	Current	Change
Wasm	(none)	13.84 MiB	+100%
JS	(none)	109.66 kiB	+100%
arkit_scenes.rrd	(none)	40.35 MiB	+100%
detect_and_track_objects.rrd	(none)	55.22 MiB	+100%
dicom_mri.rrd	(none)	63.36 MiB	+100%
dna.rrd	(none)	0.62 MiB	+100%
human_pose_tracking.rrd	(none)	55.58 MiB	+100%
plots.rrd	(none)	0.19 MiB	+100%
structure_from_motion.rrd	(none)	6.83 MiB	+100%