Fix a minor potential issue when rebatching for GpuArrowEvalPythonExec (

NVIDIA#5995)

There is a chance that the row number of the combined batches is
just a multiple of batch size, which results in returning this huge
ColumnBatch directly.

Signed-off-by: Bobby Wang <wbo4958@gmail.com>

sql-plugin/src/main/scala/org/apache/spark/sql/rapids/execution/python/GpuArrowEvalPythonExec.scala

@@ -160,7 +160,7 @@ class RebatchingRoundoffIterator(
 
     val rc: Long = combined.numRows()
 
-    if (rc % targetRoundoff == 0 || rc < targetRoundoff) {
+    if (rc <= targetRoundoff) {
       return combined
     }