Use ByteBuffer.allocateDirect() instead of ByteBuffer.allocate().

ByteBuffer.allocateDirect() behaves like Java NIO MappedByteBuffers in that
it consumes address space, which counts against our virtual memory allocation;
but cannot be destroyed or otherwise freed.  This was definitely contributing
to the LSF failures that I was seeing, but I'm not yet convinced that it's the
sole source of these virtual memory 'leaks'.  More tomorrow as the results of
my whole exome tests start to roll in.


git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@5693 348d0f76-0448-11de-a6fe-93d51630548a

java/src/org/broadinstitute/sting/gatk/datasources/reads/BAMSchedule.java

@@ -278,7 +278,7 @@ public class BAMSchedule implements CloseableIterator<BAMScheduleEntry> {
      * @return Newly allocated byte buffer.
      */
     private ByteBuffer allocateByteBuffer(final int size) {
-        ByteBuffer buffer = ByteBuffer.allocateDirect(size);
+        ByteBuffer buffer = ByteBuffer.allocate(size);
         buffer.order(ByteOrder.LITTLE_ENDIAN);
         return buffer;
     }