Add HA for CV and Docs: zero-downtime deploys (#273)

## Summary
- Set `replicas: 2` with `maxUnavailable: 0` / `maxSurge: 1` on CV and Docs deployments so rolling updates never drop below 2 ready pods
- Add PodDisruptionBudgets (`minAvailable: 1`) to protect against node drains and cluster maintenance
- Add Fly.io cache purge step to `cv-deploy.yaml` workflow (docs already had this) so CV deploys don't serve stale cached content

## Deployment and Testing
- [ ] `argocd app diff cv` / `argocd app diff docs` from branch
- [ ] Deploy from branch: `argocd app set cv --revision feature/ha-cv-docs-zero-downtime && argocd app sync cv`
- [ ] Verify 2 pods running: `kubectl get pods -n cv --context=minikube-indri`
- [ ] Test rolling restart: `kubectl rollout restart deployment/cv -n cv --context=minikube-indri`
- [ ] During rollout, confirm continuous availability via `curl -I https://cv.eblu.me`
- [ ] After merge: reset ArgoCD to main, re-sync both apps

Reviewed-on: https://forge.ops.eblu.me/eblume/blumeops/pulls/273

argocd/manifests/docs/deployment.yaml

@@ -5,7 +5,12 @@ metadata:
   name: docs
   namespace: docs
 spec:
-  replicas: 1
+  replicas: 2
+  strategy:
+    type: RollingUpdate
+    rollingUpdate:
+      maxUnavailable: 0
+      maxSurge: 1
   selector:
     matchLabels:
       app: docs

argocd/manifests/docs/kustomization.yaml

@@ -6,6 +6,7 @@ resources:
   - deployment.yaml
   - service.yaml
   - ingress-tailscale.yaml
+  - pdb.yaml
 images:
   - name: registry.ops.eblu.me/blumeops/quartz
     newTag: v1.28.2-ffa8727

argocd/manifests/docs/pdb.yaml

@@ -0,0 +1,10 @@
+---
+apiVersion: policy/v1
+kind: PodDisruptionBudget
+metadata:
+  name: docs
+spec:
+  minAvailable: 1
+  selector:
+    matchLabels:
+      app: docs