[spark][bug] Batch read PK table failed when use random column as primary key

[Yohahaha]

I found a bug while testing reading PK tables, it fails when using the last column as the primary key, current cases in SparkPrimaryKeyTableReadTest all use first column and partition column as primary key.

test("Spark Read: primary key table with last pk") {
    withTable("t") {
      sql("CREATE TABLE t (id int, name string, pk int, pk2 string) TBLPROPERTIES('primary.key'='pk,pk2')")
      checkAnswer(sql("SELECT * FROM t"), Nil)
      sql("INSERT INTO t VALUES (1, 'a', 10, 'x'), (2, 'b', 20, 'y')")
      checkAnswer(sql("SELECT * FROM t ORDER BY id"), Row(1, "a", 10, "x") :: Row(2, "b", 20, "y") :: Nil)
    }
  }

above case will failed with

Job aborted due to stage failure: Task 0 in stage 3.0 failed 1 times, most recent failure: Lost task 0.0 in stage 3.0 (TID 4) (192.168.0.116 executor driver): java.lang.ArrayIndexOutOfBoundsException: Index 2 out of bounds for length 2
	at org.apache.fluss.row.ProjectedRow.getInt(ProjectedRow.java:90)
	at org.apache.fluss.row.InternalRow.lambda$createFieldGetter$ff31e09f$6(InternalRow.java:198)
	at org.apache.fluss.row.encode.CompactedKeyEncoder.encodeKey(CompactedKeyEncoder.java:83)
	at org.apache.fluss.spark.read.FlussUpsertPartitionReader$$anon$1.compare(FlussUpsertPartitionReader.scala:113)
	at org.apache.fluss.spark.read.FlussUpsertPartitionReader$$anon$1.compare(FlussUpsertPartitionReader.scala:111)
	at org.apache.fluss.spark.utils.LogChangesIterator.hasSamePrimaryKey(LogChangesIterator.scala:117)
	at org.apache.fluss.spark.utils.LogChangesIterator.hasNext(LogChangesIterator.scala:85)
	at org.apache.fluss.client.table.scanner.SortMergeReader.readBatch(SortMergeReader.java:90)
	at org.apache.fluss.spark.read.FlussUpsertPartitionReader.initialize(FlussUpsertPartitionReader.scala:217)
	at org.apache.fluss.spark.read.FlussUpsertPartitionReader.<init>(FlussUpsertPartitionReader.scala:86)
	at org.apache.fluss.spark.read.FlussUpsertPartitionReaderFactory.createReader(FlussPartitionReaderFactory.scala:61)

Originally posted by @Yohahaha in #2523 (comment)