Change GeoIpCache and EnrichCache to LongAdder

[szybia]

From the javadocs:

[LongAdder] is usually preferable to AtomicLong when multiple threads update a common sum that is used for purposes such as collecting statistics, not for fine-grained synchronization control.

Since we largely write stats and read very infrequently, let's avoid any lock-contention possibilities by switching to the more appropriate data type.

[elasticsearchmachine]

Pinging @elastic/es-security (Team:Security)

[elasticsearchmachine]

Hi @szybia, I've created a changelog YAML for you.

[joegallo]

There's more prior art (not a ton, but some) for using a CounterMetric for tracking times in the Elasticsearch codebase than for using a LongAdder directly.

joegallo@simulacron:~/Code/elastic/elasticsearch $ git grep 'new LongAdder' | grep -i time | grep -i -E '(nanos|millis)' | grep -iv test/
x-pack/plugin/searchable-snapshots/src/main/java/org/elasticsearch/xpack/searchablesnapshots/store/input/DirectBlobContainerIndexInput.java:            private final LongAdder timeNanos = new LongAdder();
joegallo@simulacron:~/Code/elastic/elasticsearch $ git grep 'new CounterMetric()' | grep -i time | grep -i -E '(nanos|millis)' | grep -iv test/
server/src/main/java/org/elasticsearch/index/engine/Engine.java:        private final CounterMetric throttleTimeMillisMetric = new CounterMetric();
server/src/main/java/org/elasticsearch/ingest/IngestMetric.java:    private final CounterMetric ingestTimeInNanos = new CounterMetric();
server/src/main/java/org/elasticsearch/repositories/blobstore/BlobStoreSnapshotMetrics.java:    private final CounterMetric snapshotRateLimitingTimeInNanos = new CounterMetric();
server/src/main/java/org/elasticsearch/repositories/blobstore/BlobStoreSnapshotMetrics.java:    private final CounterMetric restoreRateLimitingTimeInNanos = new CounterMetric();
server/src/main/java/org/elasticsearch/repositories/blobstore/BlobStoreSnapshotMetrics.java:    private final CounterMetric uploadTimeInMillis = new CounterMetric();
server/src/main/java/org/elasticsearch/repositories/blobstore/BlobStoreSnapshotMetrics.java:    private final CounterMetric uploadReadTimeInNanos = new CounterMetric();
x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/action/TransportGetTrainedModelsStatsAction.java:        CounterMetric ingestTimeInMillis = new CounterMetric();

Is there any overriding reason to be unique here, or does it make more sense to just do what we usually do?

[joegallo]

I'd be fine with +1ing the type changes by themselves, but in order to change the logic on the time math, I'm going to have to start by understanding the reason this fence was here in the first place.

[elasticsearchmachine]

Pinging @elastic/es-data-management (Team:Data Management)

[elasticsearchmachine]

Hi @szybia, I've updated the changelog YAML for you.

[elasticsearchmachine]

💚 Backport successful

Status	Branch	Result
✅	8.18
✅	8.19
✅	9.1
✅	9.0