deleteVariantResources in kubernetes_multicluster plugin uses narrower workload check than the kubernetes plugin

[Ankitdotraider]

Plugin(s)

general

What happened?

While working on CronJob support for FindReferencingConfigMaps/FindReferencingSecrets, I noticed deleteVariantResources behaves differently between the two Kubernetes plugins.
In the regular kubernetes plugin, it checks if a resource is a workload using IsWorkload() this covers Deployment, StatefulSet, DaemonSet, ReplicaSet, and Pod.
In the kubernetes_multicluster plugin, the same function only checks IsDeployment() || IsStatefulSet().
So a canary/baseline resource that's a DaemonSet, ReplicaSet, or Pod won't be treated as a "workload" the same way on multicluster as it is on single cluster. It still gets deleted, but it's bucketed differently internally, which can cause it to be cleaned up in a different order than intended during rollback or canary teardown.
Not a crash, just an inconsistency between two plugins meant to behave the same way, and it could cause subtle ordering issues for anyone running DaemonSets on the multicluster plugin.

Steps to Reproduce

Set up a PipeCD Kubernetes multicluster deployment using a DaemonSet (or ReplicaSet/Pod) as the workload, with canary or baseline variants enabled.
Trigger a rollback or deployment that calls deleteVariantResources.
Compare the cleanup order/behavior against the same setup on the single cluster kubernetes plugin.

Expected Result

Both plugins should classify the same set of workload kinds (Deployment, StatefulSet, DaemonSet, ReplicaSet, Pod) when deciding cleanup order, since IsWorkload() already exists and is used elsewhere for this exact purpose.

Plugin version

built from main

Piped version

built from main

Environment information

Piped Environment

PipeCD Control Plane version: built from main (commit 1f2e3fbae)
OS: macOS (Ankit's local dev environment)

Piped configuration

apiVersion: pipecd.dev/v1beta1
kind: Piped
spec:
  plugins:
    - name: kubernetes_multicluster
      port: 7001
      url: https://xxxxxx
      config:
        ...
  deployTargets:
    - name: xxx-dev
      config:
        ...

Application configuration

apiVersion: pipecd.dev/v1beta1
kind: Application
spec:
  pipeline:
    ...
  plugins:
    - name: kubernetes_multicluster
      ...

Log output

Not applicable, this issue was found via code review (comparing 
deleteVariantResources in the kubernetes vs kubernetes_multicluster 
plugins), not from a live deployment or runtime failure. No error logs 
are involved. the bug is a logic/behavior inconsistency between the 
two plugin implementations.

Additional context

Found this while working on CronJob support in FindReferencingConfigMaps/
FindReferencingSecrets (see #6761 / PR #7030). While comparing shared
logic between the kubernetes and kubernetes_multicluster plugins, I
noticed deleteVariantResources classifies workloads differently between
the two see "What happened" above for details.

Happy to submit a PR for this if the fix (swapping the check for
IsWorkload()) sounds right to a maintainer.