Refactor MaxImageDimensionSubscriber

[emilk]

What

• Part of Heuristic stacks video frames rather than giving them their own 2d views #7821
• Bonus: better profiling of video loading

The change is that MaxImageDimensionSubscriber parses each .mp4 to figure out its size…
…except I disabled that check last-second because we don't have access to the VideoCache, and so this PR would lead to each video getting parsed twice.

Still, I would like to get this PR merged as a necessary refactor and generally improved structure.

This is where we're heading:

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 the web demo (if applicable):
  • Using examples from latest main build: rerun.io/viewer
  • Using full set of examples from nightly build: rerun.io/viewer
• The PR title and labels are set such as to maximize their usefulness for the next release's CHANGELOG
• If applicable, add a new check to the release checklist!
• If have noted any breaking changes to the log API in CHANGELOG.md and the migration guide

• PR Build Summary
• Recent benchmark results
• Wasm size tracking

To run all checks from main, comment on the PR with @rerun-bot full-check.

[emilk]

Actually, I just realized this means each video is parsed twice, which is… not great. I'll look into giving MaxImageDimensionSubscriber access to VideoCache

[emilk]

This is the core of it. I'll work on having access to VideoCache here, and then we can resolve this in a follow-up PR.

…or we add functionality to re_mp4 to very quickly parse an .mp4 to only look for its size info 🤔

That would be easy if we had:

• Optimize mp4 parse times, and avoid duplicating their RAM use #7481

(…and used memory-mapping)