test(kube): tolerate IsNotFound on namespace cleanup

[mbevc1]

AfterTest lists namespaces by label, then deletes each. A namespace from a previous test could still be in Terminating state when listed, then finish terminating before the Delete lands, returning NotFound. This caused the flaky failure in TestGetPodsDataWithThrottling on PR #858 (it shifted timing enough to expose it).

Skip NotFound errors on cleanup deletes — the namespace is gone, which is the intent.

[claude]

Claude finished @mbevc1's task in 1m 30s —— View job

---

PR Review: test(kube): tolerate IsNotFound on namespace cleanup

• Phase 1: Static analysis (go build, go vet, go mod tidy)
• Phase 2: Code review
• Phase 3: Post review feedback

---

Static analysis: All clear — go build ./..., go vet ./..., and go mod tidy all pass with no issues.

Review: This is a clean, correct fix for a real race condition in test cleanup.

The problem is well-described: AfterTest lists namespaces by label, iterates, and deletes each one. A namespace already in Terminating state can disappear between the List and the Delete, causing a spurious NotFound error that fails an unrelated test. Guarding with apierrors.IsNotFound is the idiomatic Kubernetes pattern for this exact scenario.

The change is minimal (4 lines), correctly scoped to test infrastructure only, and doesn't affect production code. The continue skips only NotFound — all other errors still hit require.NoErrorf and fail loudly. No notes.

LGTM 👍