[FEATURE] Workspace Pod Metrics

Certification

• I certify I am an Epic Owner for Kubeflow Notebooks 2.0 and expected to create planning-related issues.

Motivation

Bella needs to see how her workspaces are consuming CPU and memory to optimize resource allocation
for her models and experiments. Without this visibility, she's flying blind — she has no way to
know if she requested too many resources (wasting cluster capacity) or too few (risking OOM kills
or throttling) until something breaks.

Joel needs resource utilization visibility to identify capacity issues before they impact users.
When Bella reports "my workspace is slow," Joel's first question is whether the workspace is
resource-constrained — and today he has to leave the Notebooks UI and use separate tooling to
answer that.

This feature introduces resource usage metrics to the workspace UI, querying a suitable Kubernetes
metrics source and displaying point-in-time CPU and memory usage in the workspace detail view.
When no metrics source is available, the UI degrades gracefully with a clear explanation.

High Level Design / Mock-ups

Key constraints guiding the design:

• Metrics source must not impose burdensome infrastructure requirements on consumers adopting
  Notebooks v2
• Point-in-time metrics are the starting point; historical/time-series data can layer on later
  (see Epic #884)
• Graceful degradation is required — workspaces must be fully functional without a metrics source
• Pod-level metrics only (node-level and cluster-aggregate metrics are out of scope)

Acceptance Criteria

• Backend provides a mechanism to query workspace pod resource usage (CPU, memory)
• Backend detects metrics source availability and returns clear status when unavailable
• Frontend displays resource usage with request/limit context in the workspace detail view
• Frontend gracefully handles the metrics source not being available
• Metrics display does not block or slow down workspace page loading
• Solution does not require consumers to deploy additional infrastructure beyond what is
  commonly already available in Kubernetes clusters
• Unit tests cover metrics retrieval, availability detection, and error handling