fix: stop config retry for stopped components

[im0x0ing]

Fix repeated configuration retry for stopped components.

When a component is stopped in 0.9, ClusterComponent.spec.stop=true is set while spec.replicas may remain greater than 0. The underlying InstanceSet is scaled to 0 and no pods exist. However, the configuration sync policy only checked the target replicas and treated pods=0 as an abnormal state, causing repeated retries and noisy logs.

Changes

• Add stopped-state detection for reconfigure params:
  • stop=true means the component is stopped in the current 0.9 mechanism.
  • replicas=0 is kept as backward compatibility for the legacy stop mechanism.
• Update sync dynamic reload policy:
  • If no pods are found and the target component is stopped, return success instead of retrying.
  • If no pods are found but the component is not stopped, keep the original retry behavior.
• Add tests for no-pod sync behavior:
  • stop=true + replicas>0 + pods=0 should not retry.
  • replicas=0 + pods=0 should not retry.
  • replicas>0 + not stopped + pods=0 should still retry.