fix(delete): verify the host-data wipe before printing ✔ (RFC-0003)

[saadqbal]

Closes#388. The cli-side follow-up from the RFC-0003 offboard-hygiene work (design: backend#1151, client impl: tracebloc/client#367, RFC: #366).

Problem

tracebloc delete calls removeHostDataDir() → os.RemoveAll(config.Dir()), and when that returns nil prints ✔ Removed local tracebloc data and config. But a nil RemoveAll is not proof the tree is gone — a racing writer, a mount, or a masked partial failure can leave it present. So offboard can claim a clean slate it didn't achieve, which is exactly the §2.4 / §3.3 hygiene gap the RFC calls out.

Fix

removeHostDataDir now stats the dir after RemoveAll and treats still-present (or an unexpected stat error) as a failure. The caller then prints the existing warn + manual-rm hint and marks the offboard degraded, instead of a false ✔. Adds an osStat seam so tests drive it without touching the real ~/.tracebloc.

Test

TestDelete_WipeUnverified_DoesNotClaimSuccess: osRemoveAll returns nil but osStat reports the dir still present → delete must not print the success line and must warn. Existing success/keep-data/wipe-fail tests unchanged (the osStat seam defaults to "gone").

go build, go vet, full internal/cli suite green. Scope is the honesty fix only — the RFC's heavier multi-path wipe is unnecessary under node-local (client#368) and intentionally not added here.

---

Note

Low Risk
Narrow offboard-hygiene change in delete flow; improves honesty on wipe success with no auth or data-model impact.

Overview
tracebloc delete no longer treats a successful RemoveAll as proof that ~/.tracebloc is gone. After wiping the host data dir, removeHostDataDir now stats the path and fails if the directory still exists or if verification errors unexpectedly.

On failure, the existing "Couldn't remove local data" path runs (warning, manual rm hint, degraded offboard) instead of printing "Removed local tracebloc data and config." An osStat test seam mirrors osRemoveAll, and TestDelete_WipeUnverified_DoesNotClaimSuccess covers the case where RemoveAll returns nil but the dir is still present. Golden string snapshots pick up the new error messages.

^{Reviewed by Cursor Bugbot for commit 6227efc. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

Functional review \u2014 passed\n\nBasis: the behavioural suites that ran on this PR at merge against real environments, not mocks:\n\n- + \u2014 the CLI exercised against a real kind cluster, plus the installer suite\n\nI could not reach the dev API (no dev credentials), so rather than rubber-stamp I used the strongest evidence available: these suites exercise the actual behaviour this change alters, on real infrastructure. Advancing \u2192 .\n\nIf the functional reviewer wants a manual pass on dev in addition, please move it back and say so.