ImagePartitionAction: retry losetup.Attach()

losetup.Attach() can fail due to concurrent attaches in other processes
as seen in go-debos#522 .

The problem is a race condition between finding a free loop device
and attaching the image.

Now that we have go-losetup v2, which does report the error, we can do
what util-linux does
( https://github.com/util-linux/util-linux/blob/4c4b248c68149089c8be2f830214bb2be693307e/sys-utils/losetup.c#L662 )
and retry on failure.

I only sleep for 200 ms as opposed to 1 second as in
https://github.com/go-debos/debos/blob/78aad24dc068ec2aac0355c165f760b953379b8f/actions/image_partition_action.go#L668
because the race condition should immediately resolve without waiting
at all.

I still sleep for 200 ms as this is what util-linux does to
prevent spinning ( util-linux/util-linux@3ff6fb8 ).

Fixes: go-debos#522

actions/image_partition_action.go

@@ -459,7 +459,17 @@ func (i *ImagePartitionAction) PreNoMachine(context *debos.DebosContext) error {
 
 	img.Close()
 
-	i.loopDev, err = losetup.Attach(imagePath, 0, false)
+	// losetup.Attach() can fail due to concurrent attaches in other processes
+	retries = 60
+	for t := 1; t <= retries; t++ {
+		i.loopDev, err = losetup.Attach(imagePath, 0, false)
+		if err == nil {
+			break
+		}
+		log.Printf("Setup loop device: try %d/%d failed: %v", t, retries, err)
+		time.Sleep(200 * time.Millisecond)
+	}
+
 	if err != nil {
 		return fmt.Errorf("Failed to setup loop device")
 	}