[SPARK-43378][CORE] Properly close stream objects in deserializeFromC…

…hunkedBuffer

### What changes were proposed in this pull request?

Fixes a that SerializerHelper.deserializeFromChunkedBuffer does not call close on the deserialization stream. For some serializers like Kryo this creates a performance regressions as the kryo instances are not returned to the pool.

### Why are the changes needed?
This causes a performance regression in Spark 3.4.0 for some workloads.

### Does this PR introduce _any_ user-facing change?
No

### How was this patch tested?
Existing tests.

Closes #41049 from eejbyfeldt/SPARK-43378.

Authored-by: Emil Ejbyfeldt <[email protected]>
Signed-off-by: Sean Owen <[email protected]>

core/src/main/scala/org/apache/spark/serializer/SerializerHelper.scala

@@ -49,6 +49,8 @@ private[spark] object SerializerHelper extends Logging {
       serializerInstance: SerializerInstance,
       bytes: ChunkedByteBuffer): T = {
     val in = serializerInstance.deserializeStream(bytes.toInputStream())
-    in.readObject()
+    val res = in.readObject()
+    in.close()
+    res
   }
 }