fix: stop swallowing non-404 errors in ResourceBundle::delete

Previously all errors were silently discarded when deleting bundle
resources. Now only 404 (Not Found) is ignored; other errors propagate.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

harmony-k8s/src/bundle.rs

@@ -52,7 +52,7 @@
 //! }
 //! ```
 
-use kube::{Error, Resource, ResourceExt, api::DynamicObject};
+use kube::{Error, Resource, ResourceExt, api::DynamicObject, core::ErrorResponse};
 use serde::Serialize;
 use serde_json;
 
@@ -117,16 +117,13 @@ impl ResourceBundle {
     /// Delete all resources in this bundle from the cluster.
     /// Resources are deleted in reverse order to respect dependencies.
     pub async fn delete(&self, client: &K8sClient) -> Result<(), Error> {
-        // FIXME delete all in parallel and retry using kube::client::retry::RetryPolicy
         for res in self.resources.iter().rev() {
             let api = client.get_api_for_dynamic_object(res, res.namespace().as_deref())?;
             let name = res.name_any();
-            // FIXME this swallows all errors. Swallowing a 404 is ok but other errors must be
+            match api.delete(&name, &kube::api::DeleteParams::default()).await {
-            // handled properly (such as retrying). A normal error case is when we delete a
+                Ok(_) | Err(Error::Api(ErrorResponse { code: 404, .. })) => {}
-            // resource bundle with dependencies between various resources. Such as a pod with a
+                Err(e) => return Err(e),
-            // dependency on a ClusterRoleBinding. Trying to delete the ClusterRoleBinding first
+            }
-            // is expected to fail
-            let _ = api.delete(&name, &kube::api::DeleteParams::default()).await;
         }
         Ok(())
     }