Re-enable accounting breaker check in InternalTestCluster (#38131)

Relates to #30290

The intent for this is to see whether this failure still happens, and if so, provide more up-to-date logs for analysis.

Author: dakrone

test/framework/src/main/java/org/elasticsearch/test/InternalTestCluster.java

@@ -2424,10 +2424,9 @@ public void ensureEstimatedStats() {
                 final CircuitBreakerService breakerService = getInstanceFromNode(CircuitBreakerService.class, nodeAndClient.node);
                 CircuitBreaker fdBreaker = breakerService.getBreaker(CircuitBreaker.FIELDDATA);
                 assertThat("Fielddata breaker not reset to 0 on node: " + name, fdBreaker.getUsed(), equalTo(0L));
-                // TODO: This is commented out while Lee looks into the failures
-                // See: https://github.com/elastic/elasticsearch/issues/30290
-                // CircuitBreaker acctBreaker = breakerService.getBreaker(CircuitBreaker.ACCOUNTING);
-                // assertThat("Accounting breaker not reset to 0 on node: " + name, acctBreaker.getUsed(), equalTo(0L));
+                CircuitBreaker acctBreaker = breakerService.getBreaker(CircuitBreaker.ACCOUNTING);
+                assertThat("Accounting breaker not reset to 0 on node: " + name + ", are there still Lucene indices around?",
+                        acctBreaker.getUsed(), equalTo(0L));
                 // Anything that uses transport or HTTP can increase the
                 // request breaker (because they use bigarrays), because of
                 // that the breaker can sometimes be incremented from ping