Fixes #3169 : Avoid firing unneeded Events

[SamuelBussmann]

Based on #3170 Track event listeners keyed by type to allow earlier event firing veto

• Fixes #3169 : Track event listeners keyed by type to allow earlier event firing veto #3170 only triggers the listeners interested in the current event
• Filter irrelevant EventTypes already in InvocationScopedEventSink if no listener is interested in this event
• Run unordered synch listeners directly in current thread of CacheEventDispatcherImpl instead of delegating to executor service. (20us vs 1us)

This provides the following improvements:

• No deadlock on recursive cache modifications if no listener is registerd for those events
• Improved performance when listeners are used

Perf:
Testcase: 20k add(),get(),udate(),get(),remove() with a NoOp Listener on Removes

-> Baseline rev_ehcache=3.10.8

• test no_listener ended average duration was 20ms
• test unordered_asynch ended average duration was 45ms
• test unordered_synch ended average duration was 1411ms
• test ordered_asynch ended average duration was 59ms
• test ordered_synch ended average duration was 1418ms

-> MR rev_ehcache =3.10_event-firing-veto

• test no_listener average duration was 18ms
• test unordered_asynch average duration was 36ms <- 50% overhead to submit to ExecutorService?
• test unordered_synch average duration was 21ms
• test ordered_asynch average duration was 38ms
• test ordered_synch average duration was 594ms

[SamuelBussmann]

Can you retrigger the test pipelines? locally the test runs successful and it timed out on your infrastructure

[SamuelBussmann]

Hello
It's great to see the renewed activity and releases.
Is there a chance and possibly a timeline that this MergeRequest or a different fix for the original issue can be merged?