Partial all-database inspections can return a successful exit status #98

Problem and proposed fix

When at least one database succeeded, runInspectAll calculated its exit status only from successful contexts, ignoring a failed database. Preserve the successful subset of the report, but let execution failure (exit 3) take precedence over finding thresholds. The all-failed and complete-success cases retain their existing behavior.

Verification

• The actual main/Cobra subprocess regression fails on the base behavior: partial clean inspections exit 0 and partial critical inspections exit 2 rather than 3. Only discovery/collection are injected; CLI parsing, fanout, rendering and process exit remain real.
• Affected command race tests, clean committed-HEAD bash scripts/gate.sh, and go test -race ./... pass.
• Controls cover complete clean/warn/critical results, fail-on none and critical, partial failures, all failures, retained JSON output and skip diagnostics.

Scope

No collector SQL, model fields, dependencies or migrations change. PR #38 overlaps the fanout file but its partial-failure handling is conditional on all-instances; this corrects plain all-databases independently. No live connection-failure simulation is claimed. Revert this commit to restore the former exit precedence.

I would like to address this with the focused fix and regression coverage described above. Existing issues and open PRs were checked for overlap before preparing the change.