feat(ui): per-kind table columns for non-pod resources

[ErfanY]

Summary

Non-pod views previously collapsed every resource to a single loose Summary string (label count, service type, or -). This replaces it with meaningful, fixed-width, kubectl get-style columns per kind, rendered after the universal Namespace · Name · Status · Age.

Kind	Extra columns
Deployments	UP-TO-DATE AVAILABLE
ReplicaSets	DESIRED CURRENT READY
StatefulSets	READY
DaemonSets	DESIRED READY AVAIL
Services	TYPE CLUSTER-IP PORTS
Ingresses	CLASS HOSTS ADDRESS
Jobs	COMPLETIONS DURATION
CronJobs	SCHEDULE SUSPEND ACTIVE
ConfigMaps / Secrets	DATA / TYPE DATA
ServiceAccounts	SECRETS
Nodes	ROLES VERSION
PVCs / PVs	VOLUME CAPACITY ACCESS STORAGECLASS / + RECLAIM CLAIM
HPAs	REFERENCE MINPODS MAXPODS REPLICAS
PDBs	MIN-AVAIL MAX-UNAVAIL ALLOWED

Kinds without distinctive columns render just the four universal ones.

── Services ──
│default   my-resource-name   [OK] -      0s   LoadBalancer  10.0.0.5   80/TCP,443/TCP │
── PVCs ──
│default   my-resource-name   [OK] Bound  0s   pv-001        10Gi  RWO  gp3            │

Design

• ResourceKind::extra_columns() is the single source of truth for header labels + widths.
• extract_columns() in the cluster provider emits the matching row values, in the same order, at ingest (full JSON is still dropped after extraction — lean-entity model preserved).
• A contract test asserts the header count and emitted value count stay in lockstep for all 24 kinds.

Model change

summary: String → columns: Vec<String>. Pods now carry an empty vec, which drops the previously-unused per-pod node= string on the 10k-pod hot path (small net memory win). Filtering and :dump/copy span all columns.

Testing

• make validate green locally (94 tests + release build)
• new tests: column contract (all kinds), deployment/service/pvc extraction, non-pod render snapshot

🤖 Generated with Claude Code