[To dev/1.3] Fix flaky space quota result polling in cluster tests

[Caideyipi]

Description

Backport #18850 (0a79a49) to dev/1.3.

The space quota IT on dev/1.3 has the same polling helper and cluster result comparison as master. Heartbeat updates can temporarily produce different quota usage snapshots across DataNodes, causing InconsistentDataException to abort polling. Retry this exception within the existing 30-second timeout so the test can wait for quota usage to converge.

The cherry-pick applied without conflicts and changes only IoTDBSpaceQuotaIT.

Validation

• Passed mvn -B -ntp -Ddevelocity.off=true spotless:apply -pl integration-test -P with-integration-tests with no formatting changes.
• Checkstyle, Spotless, and integration-test main/test compilation passed during the targeted cluster IT build.
• Attempted mvn -o -B -ntp verify -pl integration-test -P with-integration-tests,ClusterIT -DskipUTs '-Dit.test=IoTDBSpaceQuotaIT#setSpaceQuotaTest5' -DfailIfNoTests=false -Dfailsafe.failIfNoSpecifiedTests=false. Cluster setup was blocked by repeated ConfigNode Thrift RPC NullPointerException errors (ServerContext.setRemoteAddress received a null connection context) while using locally cached 1.3.7-SNAPSHOT runtime artifacts, before the test body could execute.
• Passed git diff --check.

This PR has:

• been self-reviewed.
• added comments explaining why transient cluster inconsistency needs to be retried.