feat: consolidate event processing and optimize application performance

[jboix]

Description

Resolves #22 by updating the event processing system to improve efficiency.

Changes Made

• Switched from Reactor to Kotlin Flow for event handling to improve readability.
• Events are now saved in batch, reducing concurrency and improving performance.
• Removed reconciliation logic with a more aggressive in memory cache approach.
• Replaced Caffeine with a simple LinkedHashMap cache to support LRU eviction policy.
• Added configuration parameters to adjust the size of the cache, buffer and batch chunk size. These
  properties can also be modified with environment variables.
• Removed actuator: benchmarks are now log, reducing memory consumption and eliminating the need
  for starting a server.
• Updated dependencies to the latest compatible versions.
• Replaced dynamic mapping in OpenSearch with a handmade mapping for document fields. This change
  reduces storage requirements, and eliminates unnecessary fields, making queries more logical and
  efficient.

Checklist

• I have followed the project's style and contribution guidelines.
• I have performed a self-review of my own changes.
• I have made corresponding changes to the documentation.
• I have added tests that prove my fix is effective or that my feature works.

[github-actions]

Coverage Report

Overall Project	59.7% -31.84%	🍏
Files changed	26.45%	🍏

File	Coverage
EventDispatcherClientConfiguration.kt	100%	🍏
AliasSetupTask.kt	100%	🍏
LRUCache.kt	100%	🍏
BenchmarkScheduledLogger.kt	100%	🍏
EventRequest.kt	98.51%	🍏
OpenSearchSetupService.kt	96.7%	🍏
OpenSearchConfigurationProperties.kt	90.14%	🍏
OpenSearchConfiguration.kt	78.38% -13.51%	🍏
StatsTracker.kt	35.85% -64.15%	🍏
RetryProperties.kt	29.31% -70.69%	🍏
TimeTracker.kt	27.06% -72.94%	🍏
LoggingExtensions.kt	23.7% -18.52%	🍏
PillarboxDataTransferApplication.kt	18.18%	🍏
EventDispatcherClient.kt	10.05% -84.69%	🍏
FlowUtils.kt	0%	🍏
DataTransferApplicationRunner.kt	0% -82.42%	🍏
MovingAverageCalculator.kt	0%	🍏
ElasticsearchRepositoryExtensions.kt	0%	🍏

[ben-manes]

I'd recommend using Guava's Cache instead of Ehcache if you require LRU. I may be biased, but my opinion is that library is not safe for production users due to a trivially exploitable denial of service attack. Both have bugs and are inactive.

fwiw - Caffeine's eviction policy is adaptive, so it starts biased towards frequency and adjusts towards recency based on the observed workload. It does not provide any guarantees if you rely on LRU semantics for correct application behavior, as a more important constraint than the hit rate. If that is the case, then I think making the logic more explicit about the LRU requirement (e.g. via a LinkedHashMap or similar) is better for maintainability.

[jboix]

@ben-manes Thank you for the feedback! I've used Caffeine before and really appreciate its simplicity. However, for our particular use case, we require a strict LRU eviction policy.

After evaluating your comment, I reconsidered our previous implementation and decided to drop EhCache and add a simple LRU cache using LinkedHashMap. This approach is straightforward and avoids introducing an additional dependency, which we value.

[github-actions]

🎉 This PR is included in version 3.1.0 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀