test(data delete): cover the destructive teardown path (execute + partial-fail exit 7)

[LukasWodka]

Summary

Third increment from the CLI test-coverage audit — the destructive data delete path, a P0 (runDataDelete was 22%; every existing test stopped at cluster discovery, so the actual teardown — a DROP TABLE + file removal that can't be undone — had no coverage).

Production changes (seams — zero behavior change)

Two fn-var test seams, mirroring the existing loadClusterFn / listDatasetsFn pattern (both default to the real function, so production is unchanged):

1. resolveClusterTargetFn — lets a command test inject a fully-resolved target (fake clientset + release + PVC) instead of seeding the k8s objects discoverRelease / DiscoverSharedPVC look for. runDataDelete now calls it.
2. teardownFn — over push.Teardown (the DROP TABLE + file removal). runDataDelete's execute step now calls it.

Test

TestRunDataDelete_Execute drives the command past discovery via the seams, then exercises the three teardown outcomes:

• clean → exit 0 + a Deleted line;
• table dropped, file removal fails → exit 7 + the recovery hint (… was dropped, but removing its files failed — the idempotent-DROP re-run guidance);
• fails before the drop → exit 7 teardown failed.

The mixed-case "Churn" input in the partial-failure case also pins the case-insensitive resolveDeleteTarget match (backend#1027 — a mis-cased name used to DROP nothing and still exit 0).

Why a seam, not a k8s fixture

resolveClusterTarget's success needs discoverRelease + DiscoverSharedPVC to find seeded objects — fiddly, and it's the established seam pattern that unblocks every command that resolves a target. Testing resolveClusterTarget's own success tail (its 47%) with real k8s objects is a separate follow-up.

Not in this PR (follow-ups)

• Routing runClusterInfo / runClusterDoctor / realProbeEnv through loadClusterFn, and the stageFn seam for runDataIngest's --overwrite path — plus their tests.
• resolveClusterTarget's own success-tail coverage (needs the k8s-object fixture).

Test plan

• make ci green (build, -race, gofmt, lint, schema). New + existing data delete tests pass.

🤖 Generated with Claude Code

---

Note

Low Risk
No production behavior change; seams default to existing functions and the diff is tests plus indirection wiring only.

Overview
Adds test-only fn-var seams so runDataDelete can be exercised past cluster discovery without a real kubeconfig or cluster: resolveClusterTargetFn (injected resolved target) and teardownFn (over push.Teardown). Production still uses the real implementations; only call sites in runDataDelete switch to the seam variables.

TestRunDataDelete_Execute drives the destructive execute step with --yes and mocked targets, covering success (exit 0 + Deleted), partial failure after DROP TABLE (exit 7 + recovery hint), and failure before the drop (exit 7 teardown failed). The partial-failure case uses mixed-case Churn to lock in case-insensitive target resolution before teardown.

^{Reviewed by Cursor Bugbot for commit 1b0e871. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 1b0e871. Configure here.}

[saadqbal]

LGTM 👍 The three teardown outcomes line up exactly with runDataDelete's error handling, and pinning the case-insensitive resolveDeleteTarget match (backend#1027) via the Churn case is a nice touch. Seams default to the real fns, so no behavior change. Note: since this stack adds seams to data.go, expect a rebase as the siblings land.