[cxzl25]

Purpose

Clustering a paimon table on a TIMESTAMP column fails on Spark 3.4 and above:

Cannot use column ts of type TimestampNTZType in ZOrdering, the type is unsupported
java.lang.IllegalArgumentException: Cannot use column ts of type TimestampNTZType in ZOrdering, the type is unsupported
	at org.apache.paimon.spark.sort.SparkZOrderUDF.sortedLexicographically(SparkZOrderUDF.java:354)
	at org.apache.paimon.spark.sort.ZorderSorter.lambda$zValue$0(ZorderSorter.java:71)

Tests

Add UT

[JingsongLi]

Requirement fit: SUPPORTED (triage: GO)\nImplementation: CLEAN\n\nThis fixes an end-to-end Spark clustering failure for the TimestampNTZ type produced by current Paimon timestamp mappings. I reviewed null handling, monotonic conversion, Z-order and Hilbert dispatch, and Spark-version compatibility. The seconds-level representation matches the existing TimestampType cast path; no actionable P0/P1/P2 issue found.