threads: Bulk process thread subscription updates from sync and companion enpoint

[razvp]

Fixes #5610

• Public API changes documented in changelogs (optional)

Signed-off-by: Razvan [email protected]

[codecov]

Codecov Report

❌ Patch coverage is 94.27313% with 13 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.61%. Comparing base (10ff5d0) to head (07a7e45).
⚠️ Report is 14 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
...tes/matrix-sdk-base/src/store/integration_tests.rs	93.42%	0 Missing and 10 partials ⚠️
crates/matrix-sdk-sqlite/src/state_store.rs	94.28%	0 Missing and 2 partials ⚠️
...ates/matrix-sdk/src/client/thread_subscriptions.rs	97.22%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5848      +/-   ##
==========================================
+ Coverage   88.60%   88.61%   +0.01%     
==========================================
  Files         361      361              
  Lines      102294   102476     +182     
  Branches   102294   102476     +182     
==========================================
+ Hits        90637    90809     +172     
- Misses       7438     7439       +1     
- Partials     4219     4228       +9     

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

[codspeed-hq]

CodSpeed Performance Report

Merging #5848 will not alter performance

_{Comparing razvp:thread-subs-bulk-upsert (07a7e45) with main (a3424a7)}

Summary

✅ 50 untouched

[Hywan]

Thanks! It looks like a solid contribution, thank you. However, I wonder if you could add a bit of context and to explain what you're doing, and why, please?

[razvp]

Thanks! It looks like a solid contribution, thank you. However, I wonder if you could add a bit of context and to explain what you're doing, and why, please?

The sliding sync and back-pagination endpoints provide thread subscription updates in batches, but currently the SDK stores/updates them one by one. (MSC4308)
This PR changes that by processing them in bulk, in one database transaction.

The first commit adds the upsert_thread_subscriptions() method to the StateStore trait and implements it for the 3 storage types. The IndexeddbStateStore and MemoryStore implementations are very similar to the existing single update per transaction
The second commit updates ThreadSubscriptionCatchup::sync_subscriptions and ThreadSubscriptionCatchup::thread_subscriptions_catchup_task to actually use the bulk api.

Most of the bump_stamp logic is gathered from the existing test. (ex: if an update doesn't have a bump_stamp and the stored subscription has one, then keep the existing one)

Switching to a bulk api was suggested in this comment.