fix memory leak and inconsistent behavior in memory caches

[warriordog]

What

This PR resolves #14310 and #14311 by refactoring MemoryKVCache<T>. The new implementation supports capacity limits with Least Recently Used (LRU) eviction, and no longer requires garbage collection or manual disposal. All references to this class have been updated to specify a reasonable capacity limit.

Additionally, all cache lifetimes and capacity limits have been exposed as configuration values. This allows the limits to be adjusted by instance admins to account for available system resources, which reduces the chance of cache-related resource exhaustion. Larger instances can even increase the limits, thus improving cache hit rate and reducing churn.

Why

The linked issues pose a serious reliability issue that is affecting many instances, including my own. Resolving the memory leaks will improve stability, uptime, and performance.

Additional info (optional)

This change is upstreamed from the Sharkey fork. Please see Sharkey MR #580 for additional details and context.

Checklist

• Read the contribution guide
• Test working in a local environment
• (If needed) Add story of storybook
• (If needed) Update CHANGELOG.md
• (If possible) Add tests

[github-actions]

このPRによるapi.jsonの差分

差分はこちら

Get diff files from Workflow Page

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 20.49%. Comparing base (22c4e9d) to head (2d9d548).

Additional details and impacted files

@@             Coverage Diff              @@
##           develop   #14314       +/-   ##
============================================
- Coverage    40.07%   20.49%   -19.59%     
============================================
  Files         1524      702      -822     
  Lines       188792    98756    -90036     
  Branches      3515     1027     -2488     
============================================
- Hits         75666    20243    -55423     
+ Misses      112552    77992    -34560     
+ Partials       574      521       -53     

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

[syuilo]

Thanks for the PR!

Since it seems redundant to make everything configurable about the cache used inside the system, and since this is likely to change as development progresses, it would be better to make it possible to set some single factor that applies to the entire system, such as the “rate limit” setting in the role policy.

(This is my opinion and I am looking for your feedback)

[warriordog]

This has been replaced by #14363