Unshimmify Arm helper methods to expose as public API

dist/unshimmed-from-each-spark3xx.txt

@@ -1,3 +1,5 @@
+com/nvidia/spark/rapids/Arm*
+com/nvidia/spark/rapids/CloseableHolder*
 com/nvidia/spark/rapids/*/RapidsShuffleManager.class
 com/nvidia/spark/rapids/AvroProvider.class
 com/nvidia/spark/rapids/HiveProvider.class

docs/additional-functionality/rapids-udfs.md

@@ -93,6 +93,48 @@ closed by the RAPIDS Accelerator, so the UDF only needs to close any
 intermediate data generated while producing the final result that is
 returned.
 
+#### ARM Helper Methods
+
+For Scala, helper methods implementing the automatic resource management
+pattern are exposed via the public Spark RAPIDS API under the `Arm` object:
+
+```scala
+import com.nvidia.spark.rapids.Arm.{withResource, closeOnExcept}
+```
+
+These patterns serve different use cases:
+- `withResource`: closes the resource after the code block completes regardless
+of whether it succeeds or throws an exception. This should be used for intermediate
+resources that are unused beyond the code block.
+- `closeOnExcept`: only closes the resource if an exception occurs within the 
+code block; if the block succeeds, the resource remains open. This should be used
+if you want to continue using the resource after the code block but need to clean 
+up if an error occurs.
+
+Examples:
+
+```scala
+  override def evaluateColumnar(numRows: Int, args: ColumnVector*): ColumnVector = {
+    val nullMask = withResource(args.head.isNull()) { nulls =>  // nulls always closed after code block
+      nulls.not()
+    }
+    // ...
+  }
+```
+
+```scala
+  override def evaluateColumnar(numRows: Int, args: ColumnVector*): ColumnVector = {
+    val nullsCol = closeOnExcept(args.head.isNull()) { nulls =>  // nulls only closed if exception occurs
+      validate(nulls) // do something that might throw an exception
+      nulls
+    }
+    // ...
+  }
+```
+
+Note that these methods originated as a solution for Scala 2.12. Scala 2.13 introduced
+the [Using](https://www.scala-lang.org/api/2.13.6/scala/util/Using$.html) utility, which is an alternative to `withResource` for 2.13 builds of Apache Spark.
+
 ### Generating Columnar Output
 
 The `evaluateColumnar` method must return a `ColumnVector` of an appropriate