add an allocate method with stream in java DeviceMemoryBuffer (rapids…

…ai#7826)

Looks like this is also needed in order to remove the `CudaUtil` class in the rapids spark plugin.

@abellina @jlowe

Authors:
  - Rong Ou (https://github.com/rongou)

Approvers:
  - Jason Lowe (https://github.com/jlowe)
  - Alessandro Bellina (https://github.com/abellina)

URL: rapidsai#7826

java/src/main/java/ai/rapids/cudf/DeviceMemoryBuffer.java

@@ -122,7 +122,17 @@ private DeviceMemoryBuffer(long address, long lengthInBytes, DeviceMemoryBuffer
    * @return the buffer
    */
   public static DeviceMemoryBuffer allocate(long bytes) {
-    return Rmm.alloc(bytes);
+    return allocate(bytes, Cuda.DEFAULT_STREAM);
+  }
+
+  /**
+   * Allocate memory for use on the GPU. You must close it when done.
+   * @param bytes size in bytes to allocate
+   * @param stream The stream in which to synchronize this command
+   * @return the buffer
+   */
+  public static DeviceMemoryBuffer allocate(long bytes, Cuda.Stream stream) {
+    return Rmm.alloc(bytes, stream);
   }
 
   /**