fix(ci): workflow path verification, stale legacy path cleanup, integration smoke test

[amar-python]

CI & Infrastructure Stability (High Priority)

Based on feature/schema-validation (PR #7) since that branch contains the restructured backend/ layout these fixes target. Merging PR #7 first will bring these changes into main cleanly.

1. Workflow Path Verification

• New tools/verify_workflow_paths.py — stdlib-only checker that parses every .github/workflows/*.yml, extracts file/directory references from run: commands and local uses: actions, and exits non-zero if any reference no longer exists in the working tree. Prevents the "repo restructured but CI still points at old paths" failure class permanently.
• New validation CI job staged in docs/ci/quality-gate.yml that runs the checker on every push/PR.
  • ⚠️ Maintainer action needed: the CI bot has no workflows permission, so the updated workflow is staged in docs/ci/quality-gate.yml (the repo's existing convention). Please copy it to .github/workflows/quality-gate.yml when merging.

2. Global Path Update (stale legacy roots → backend/migration/…)

Fixed all remaining references to the pre-restructure locations (tests/, build/, evals/):

• scripts/eval_list.py, scripts/eval_summary.py, scripts/eval_compare.py — REPORTS_DIR pointed at the non-existent root evals/reports/; now backend/migration/evals/reports/ (matches where runner.py actually writes)
• scripts/run_qa.ps1 — all pytest/coverage/lint targets updated
• scripts/test.sh, scripts/test.ps1 — layer 2/3 runner paths updated
• preflight.sh, preflight.ps1 — project-root existence checks updated
• Correctly-pointing paths (Makefile, lint.sh, health_check.py, select_tests.py) were left untouched.

3. Integration Smoke Test

• New backend/tests/integration/test_smoke.py — asserts GET /api/health returns 200 OK (plus payload shape and root-route checks), marked @pytest.mark.integration, runs on the existing SQLite-backed pytest fixtures.

Verification

• python3 tools/verify_workflow_paths.py --verbose → 8 references checked, all resolve ✔ (and correctly fails on a synthetic stale workflow)
• pytest backend/tests/ → 41 passed (38 existing + 3 new)
• bash -n clean on all touched shell scripts; preflight.sh passes its project-file checks
• flake8 clean on the new tool

Checklist

• Tests added
• Verified workflow paths