Support null equal in last join for openmldb-batch

We may support `===` and `<=>` for last join.

```
--- a/java/openmldb-batch/src/main/scala/com/_4paradigm/openmldb/batch/nodes/JoinPlan.scala
+++ b/java/openmldb-batch/src/main/scala/com/_4paradigm/openmldb/batch/nodes/JoinPlan.scala
@@ -91,10 +91,18 @@ object JoinPlan {
for (i <- 0 until keyNum) {
val leftColumn = SparkColumnUtil.resolveExprNodeToColumn(leftKeys.GetChild(i), node.GetProducer(0), leftDf)
val rightColumn = SparkColumnUtil.resolveExprNodeToColumn(rightKeys.GetChild(i), node.GetProducer(1), rightDf)
- joinConditions += (leftColumn === rightColumn)
+ joinConditions += leftColumn.eqNullSafe(rightColumn)
}
}
```

The change is in JoinPlan.scala, specifically the join condition logic. Start by understanding the existing last join implementation and how null equality is currently handled. Look at the SparkColumnUtil and the eqNullSafe method. Run tests related to join operations to verify the new behavior.

Vom Indexierungsmodell aus dem Issue-Text verfasst.