fix(data list): route through the listDatasetsFn seam + cover the query path

[LukasWodka]

Summary

Second increment from the CLI test-coverage audit — the data list path. A focused slice of the "test seams" work: fixes a genuine consistency bug in production and makes ListDatasets's query path testable, with the tests.

Production changes

1. runDataList now routes through the listDatasetsFn seam (data_list.go:118). It was calling push.ListDatasets directly, while the listDatasetsFn seam it should use already exists and is used two lines over by destTableExists / resolveDeleteTarget (data.go:1462). This was a real consistency bug — data list was the only dataset-query caller bypassing the seam — and it left runDataList's success/error paths untestable.

2. ListDatasets split so its exec path is fakeable. The function built a real SPDYExecutor internally, so the exec + error-wrap + parse logic sat at 0% (a real cluster was the only way in). Extracted listDatasetsWith(ctx, exec Executor, ns, pod, container) — ListDatasets still locates the mysql pod and builds the production executor, then delegates. Public signature and the listDatasetsFn seam are unchanged.

Test changes

• TestListDatasetsWith (new) — drives the split-out core with a fake Executor: a query result is parsed into table names, an empty database yields no datasets, and a failing exec surfaces querying datasets + the remote stderr (via stderrSuffix).
• fakeExecutor gains a stdoutToReturn field (stream_test.go) — it previously discarded the stdout writer (it was built for tar-streaming, which doesn't read stdout). Now it can return a stdout body, which ListDatasets needs. Existing callers don't set it → no behavior change (all push tests still pass).

Not in this PR (follow-ups)

The heavier seam work stays separate so each PR is reviewable:

• Routing runClusterInfo / runClusterDoctor / realProbeEnv through loadClusterFn/newClientsetFn, and the teardownFn/stageFn seams — plus the command-path tests they unblock (runDataDelete execute + partial-fail, runClusterInfo post-discovery, runDataList success path). Those interlock with the resolveClusterTarget-success / Bound-PVC fixture the audit flagged, which I'll build next.

Test plan

• make ci green (build, -race, gofmt, lint, schema).
• Whole internal/push package re-run to confirm the fakeExecutor change didn't regress the tar-stream / stage / teardown tests.

🤖 Generated with Claude Code

---

Note

Low Risk
Refactor and test seams only; production behavior for listing datasets is unchanged aside from consistent use of the existing injectable hook.

Overview
tracebloc data list now calls listDatasetsFn instead of push.ListDatasets directly, matching destTableExists / resolveDeleteTarget and enabling CLI tests to stub dataset listing.

In internal/push, ListDatasets still discovers the mysql pod and builds SPDYExecutor, but the information_schema query, error wrapping, and parsing live in injectable listDatasetsWith. TestListDatasetsWith covers success, empty output, and exec failure (including stderr). fakeExecutor can return stdout for those tests without changing tar-stream behavior when stdoutToReturn is unset.

^{Reviewed by Cursor Bugbot for commit 398b293. 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 398b293. Configure here.}

[saadqbal]

Nice — good catch that data list was the only dataset-query caller skipping the listDatasetsFn seam; routing it through matches destTableExists/delete and unblocks the tests. Refactor is behavior-preserving. LGTM 👍