refactor: Add profile cache for handle_remote_event

[jmartinesp]

This should slightly improve the performance for loading timeline items since we might skip loading the sender profiles for some/most items.

In my local tests on Android, this can save up to 50ms for each timeline initialization, although the benchmark only shows a 1.7% improvement 🥲 . That said, it also says the the function that populates the cache is not called when it definitely is,
so it doesn't seem too reliable.

• Public API changes documented in changelogs (optional)

Signed-off-by:

[codecov]

Codecov Report

❌ Patch coverage is 94.11765% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 88.54%. Comparing base (781df55) to head (04902ac).
⚠️ Report is 23 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
...dk-ui/src/timeline/controller/state_transaction.rs	91.66%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5827   +/-   ##
=======================================
  Coverage   88.54%   88.54%           
=======================================
  Files         361      361           
  Lines      101421   101434   +13     
  Branches   101421   101434   +13     
=======================================
+ Hits        89803    89818   +15     
+ Misses       7410     7409    -1     
+ Partials     4208     4207    -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[codspeed-hq]

CodSpeed Performance Report

Merging #5827 will not alter performance

_{Comparing refactor/add-profiles-cache-to-handle-remote-event (04902ac) with main (781df55)}

Summary

✅ 50 untouched

[Hywan]

Thank you for the contribution!
How do you manage the user profile invalidation? If a user has updated its profile, how is it updated from the cache (the HashMap)?

[jmartinesp]

That's the neat part, you don't! Jokes aside, when I discussed it with @poljar (and I think I confirmed it by looking at the code), by the time we reach this part of the flow we'd have already applied any changes caused by the state events, including the room member ones, so it should be fine to not invalidate the cache, since it'll be used just for the current vector diffs.

I may be wrong though.

[Hywan]

The cache is local to handle_remote_events_with_diffs. I think it's correct. Let's go :-].