Add flaky-test ranking metric and dashboard panel

[Copilot]

This change adds a Quality KPI widget that ranks the top tests by failure-flip rate across the selected dashboard window, using existing external_case_results data. It introduces a new aggregated backend endpoint and connects it to a sortable frontend table to surface flapping tests at a glance.

• Backend: flaky ranking aggregation + API

  • Added FlakyTestEntry and FlakyRankingResponse schemas.
  • Added get_flaky_ranking(db, window_days, top_n) in quality metrics CRUD.
    • Reuses existing window helpers and case-row loader.
    • Groups by stable identity (test_case_id → external_id → id).
    • Computes runs, flaky_outcomes, transitions, flip_rate, last_outcome, last_seen_at.
    • Excludes non-signal outcomes (started, skipped, aborted) from transition sequence.
    • Applies ranking sort: flip_rate desc, flaky_outcomes desc, runs desc, display_name asc.
    • Returns synthetic/empty-state reasons aligned with existing quality-metrics behavior.
  • Added GET /quality-metrics/flaky-ranking with window and top_n query params.

• Backend tests

  • Added dedicated flaky-ranking test module covering:
    • empty DB synthetic response,
    • high-flip sequence behavior (pass↔fail),
    • exclusion of all-pass and single-run tests,
    • top_n capping,
    • independent counting of flaky_outcomes,
    • window-boundary exclusion,
    • external-id-only identities with test_case_id = null.

• Frontend: API + dashboard widget

  • Added getFlakyRanking({ window, topN }) and corresponding TS types.
  • Added FlakyRankingPanel to QualityDashboardPage, driven by existing window state.
  • Renders sortable table with columns:
    • Test, Runs, Flips, Flip rate, Last outcome, Last seen.
  • Reuses existing outcome chip color mapping style.
  • Handles loading/empty/synthetic states with reason messaging.
  • Links test cells to filtered test-case view (/test-cases?search=...).

• Frontend tests

  • Added panel tests for:
    • rendering rows from mocked response,
    • synthetic empty state,
    • refetch on window change,
    • client-side sort via column headers.

# backend/app/crud/quality_metrics.py (core calculation)
considered_outcomes = [
    case_result.outcome
    for case_result, _test_case in ordered_rows
    if case_result.outcome in {CaseStatus.passed, CaseStatus.failed, CaseStatus.flaky}
]
transitions = sum(
    1 for previous, current in zip(considered_outcomes, considered_outcomes[1:], strict=False)
    if previous != current
)
flip_rate = transitions / max(len(considered_outcomes) - 1, 1)

[bg-playground]

@copilot
CI checks are still failing - See screen shot for E2E test CI

[Copilot]

@copilot
CI checks are still failing - See screen shot for E2E test CI
...

Fixed in commit fix: update QualityDashboardPage to use filters.window for FlakyRankingPanel.

Root cause: the base branch (main) had refactored selectedWindow → filters.window (with URL/localStorage persistence via useSearchParams). Our branch still used selectedWindow, so the CI merge result had <FlakyRankingPanel window={selectedWindow} /> at line 542 where selectedWindow no longer existed.

The fix rebases our page file onto main's refactored version and updates the FlakyRankingPanel prop to window={filters.window}.