perf: precompute the room notification modes only if push rules have changed

[bnjbvr]

The room notification modes will only change when the push rules have changed; otherwise, recomputing them is a no-op, and only wastes time.

Here's a distribution of the worst processing times under this method, as measured in multiverse with my personal account running (fresh, i.e. after an initial response) for a few minutes:

Before:

0.001358279s
0.019668889s
0.030104444s
0.030376296s
0.048609819s
0.049683619s
0.054289813s
0.063690030s

After:

0.000000385s
0.000000387s
0.000000410s
0.000000430s
0.000000438s
0.000000478s
0.000000487s
0.000000568s
0.010011267s

The new distribution makes sense: the push rules event comes only once, in the initial response, so they're processed only once. Then, there's no need to process it ever again (in my session I've never changed the processing times), so it's almost instantaneous the subsequent times.

Found thanks to #5612

[codecov]

Codecov Report

❌ Patch coverage is 76.92308% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.87%. Comparing base (7767ef6) to head (5ea16b9).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
crates/matrix-sdk/src/sliding_sync/client.rs	76.92%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5621      +/-   ##
==========================================
- Coverage   88.88%   88.87%   -0.01%     
==========================================
  Files         348      348              
  Lines       96451    96460       +9     
  Branches    96451    96460       +9     
==========================================
  Hits        85730    85730              
- Misses       6685     6693       +8     
- Partials     4036     4037       +1     

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

[codspeed-hq]

CodSpeed Performance Report

Merging #5621 will not alter performance

_{Comparing bnjbvr/process-room-notification-changes-only-when-push-rules-changed (5ea16b9) with main (bcabf1b)}

Summary

✅ 49 untouched benchmarks

[bnjbvr]

Note: an extra perf improvement would be to create the NotificationSettings instance once (instead of once per call to user_defined_notification_mode()), and use it to update all the room notification modes at once, instead of creating the NotificationSettings (which loads the push rules from DB) once per room.

Also, I think it's still necessary to figure what are the newly joined rooms, and compute the user defined notification mode for those.

Not as simple as I imagined first, unfortunately.

[andybalaam]

Great, thanks. I am assuming this is already covered by adequate tests?