fix: add retry and force flags to test framework DeleteDataset

The test framework's DeleteDataset was calling pool.dataset.delete with
bare params and no retry, causing EBUSY failures during NFS test cleanup
when TrueNAS hasn't released the dataset mount yet. Now matches the
driver's approach: recursive=true, force=true, with 12x5s retry on busy.

Author: fenio

tests/e2e/framework/truenas.go

@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"time"
 
+	"github.com/fenio/tns-csi/pkg/retry"
 	"github.com/fenio/tns-csi/pkg/tnsapi"
 	"k8s.io/klog/v2"
 )
@@ -107,14 +108,18 @@ func (v *TrueNASVerifier) NVMeOFSubsystemExists(ctx context.Context, nqn string)
 	return len(subsystems) > 0, nil
 }
 
-// DeleteDataset deletes a dataset from TrueNAS.
-// This is used for cleaning up retained datasets after tests.
+// DeleteDataset deletes a dataset from TrueNAS with recursive+force flags and retry logic.
+// This matches the driver's DeleteDataset approach: passes recursive=true, force=true,
+// and retries on EBUSY errors (12 attempts × 5s interval = ~60s total).
 func (v *TrueNASVerifier) DeleteDataset(ctx context.Context, datasetPath string) error {
-	var result any
-	if err := v.client.Call(ctx, "pool.dataset.delete", []any{datasetPath}, &result); err != nil {
-		return fmt.Errorf("failed to delete dataset %s: %w", datasetPath, err)
-	}
-	return nil
+	return retry.WithRetryNoResult(ctx, retry.DeletionConfig("DeleteDataset("+datasetPath+")"), func() error {
+		var result any
+		params := []any{datasetPath, map[string]any{"recursive": true, "force": true}}
+		if err := v.client.Call(ctx, "pool.dataset.delete", params, &result); err != nil {
+			return fmt.Errorf("failed to delete dataset %s: %w", datasetPath, err)
+		}
+		return nil
+	})
 }
 
 // deleteResourceByFilter is a helper that queries for a resource by filter, gets its ID, and deletes it.