fix(k8s): fix rollout status check for Recreate

Before this fix, the rollout status of deployments using the `Recreate`
strategy would be reported as ready before the new pods were fully
ready.

This was fixed by copying/following the latest logic from kubectl's
rollout status checking logic, comparing the requested number of
replicas from the deployment spec with the actual updated pod count.

See:
https://github.com/kubernetes/kubectl/blob/24d21a0ee42ecb5e5bed731f36b2d2c9c0244c35/pkg/polymorphichelpers/rollout_status.go

core/src/plugins/kubernetes/status/workload.ts

@@ -18,6 +18,7 @@ import {
   V1StatefulSetSpec,
   V1DeploymentStatus,
   CoreV1Event,
+  V1DeploymentSpec,
 } from "@kubernetes/client-node"
 import dedent = require("dedent")
 import { getCurrentWorkloadPods } from "../util"
@@ -224,8 +225,9 @@ async function getRolloutStatus(workload: Workload) {
     }
   } else {
     const status = <V1DeploymentStatus>workload.status
+    const deploymentSpec = <V1DeploymentSpec>workload.spec
 
-    const desired = status.replicas || 0
+    const desired = deploymentSpec.replicas || 1
     const updated = status.updatedReplicas || 0
     const replicas = status.replicas || 0
     const available = status.availableReplicas || 0