BUG: Fix GDCM crash when reading DICOM image

GDCM crashed when a DICOM image that had 32 bits allocated.
The problem was that DoOverlayCleanup returned failure but that return value was ignored and later an empty buffer read was attempted.

This fix avoids the crash by not ignoring the failed overlay cleanup.
Applications can detect the error and switch to DCMTK IO to read such files (DCMTK can correctly read files with 32 bits allocated).

Source/MediaStorageAndFileFormat/gdcmImageCodec.cxx

@@ -675,6 +675,7 @@ bool ImageCodec::DecodeByStreams(std::istream &is, std::ostream &os)
 
   // Do the overlay cleanup (cleanup the unused bits)
   // must be the last operation (duh!)
+  bool copySuccess = false;
   if ( PF.GetBitsAllocated() != PF.GetBitsStored()
     && PF.GetBitsAllocated() != 8 )
     {
@@ -685,21 +686,23 @@ bool ImageCodec::DecodeByStreams(std::istream &is, std::ostream &os)
     // Sigh, I finally found someone not declaring that unused bits where not zero:
     // gdcmConformanceTests/dcm4chee_unusedbits_not_zero.dcm
     if( NeedOverlayCleanup )
-      DoOverlayCleanup(*cur_is,os);
+      {
+      copySuccess = DoOverlayCleanup(*cur_is, os);
+      }
     else
       {
       // Once the issue with IMAGES/JPLY/RG3_JPLY aka gdcmData/D_CLUNIE_RG3_JPLY.dcm is solved the previous
       // code will be replace with a simple call to:
-      DoSimpleCopy(*cur_is,os);
+      copySuccess = DoSimpleCopy(*cur_is, os);
       }
     }
   else
     {
     assert( PF.GetBitsAllocated() == PF.GetBitsStored() );
-    DoSimpleCopy(*cur_is,os);
+    copySuccess = DoSimpleCopy(*cur_is, os);
     }
 
-  return true;
+  return copySuccess;
 }
 
 bool ImageCodec::IsValid(PhotometricInterpretation const &)