Optimize WeakConcurrentCollection removeAll and retainAll lookup complexity

[Smorki]

In WeakConcurrentCollection#removeAll and #retainAll, calling c.contains(value) inside the loop gets slow when c is a List.

This PR wraps c into a HashSet prior to iteration if it's not already a Set to make lookups O(1). It also handles empty collections cleanly to stay compliant with standard Java Collection behavior.

[BaconCat1]

Why did you remove Objects/NonNull imports while retaining their uses? Does not compile.

[Smorki]

oh iy was accidentally removed during the web conflict resolution. Just pushed a fix to restore the missing imports

[Smorki]

🥀

[BaconCat1]

Even after the compile issue is fixed, I don't think this is a good idea as written.
What actual Canvas workload is this optimizing? There aren't any Canvas call sites for WeakConcurrentCollection.removeAll or retainAll, so right now this adds complexity to a path that isn't even used.
It also isn't behaviorally equivalent. IdentityHashMap.values() is a non-Set collection with identity-based contains, and copying it into a HashSet changes the result for distinct but equal objects. That flips the behavior of both removeAll and retainAll.
This isn't always a performance win either. An empty or tiny receiver with a huge list still has to allocate and fill the entire HashSet, while repeated CopyOnWriteArrayList removals still dominate the worst case. Set doesn't guarantee constant-time lookups either, so the type check isn't a reliable way to choose this.
Unless there is a real call site and benchmark showing this helps, I don't think we should merge it. If there is, optimize that known path and keep collection.contains as the generic fallback.

[Smorki]

Fair points, especially regarding the behavioral difference with identity-based collections and the lack of current call sites within Canvas I see now that adding this complexity might be premature without a concrete use case or benchmark.
illl close this for now to avoid adding unnecessary overhead where its not needed. Thanks for the detailed review!