chore(ci): e2e report improve

[universal-itengineer]

Description

Improve the E2E messenger report generated by CI and reorganize the
report scripts so the same logic is easier to follow.

User-facing changes:

• Add a Reason column to the failed-tests thread table and rename
  Test group to Tests.
• Link cluster names in failed-tests thread replies to the matching
  workflow/job URL.
• Hide the Errors column in the main test-results table when there
  are no Ginkgo errors.
• Calculate Success Rate from executed specs only:
  passed / (passed + failed + errors).
• Add table header emojis to the main report.
• Use the full Ginkgo Failure.Message text as the failed-test reason.
• Stop wrapping *exec.ExitError with %w in
  test/e2e/internal/framework/ssh.go so Ginkgo no longer dumps Go
  internals into the report when an SSH command fails.
• Capture Ginkgo stdout/stderr in the reusable pipeline and parse it as
  a fallback report source so suite-level failures such as
  SynchronizedBeforeSuite are still surfaced when the JSON report is
  missing or contains only setup failure data.

Internal refactors (no behavior change, covered by existing tests):

• Move the shared withTempDir / createCore jest helpers into
  shared/test-utils.js to remove duplication across three test files.
• Extract escapeRegExp in shared/report-model.js; consolidate
  buildStatusMessage into a status → template lookup table; remove a
  redundant ternary in buildClusterStatus.
• Unify the Ginkgo source lookup (findGinkgoReport /
  findGinkgoOutput → one findGinkgoSource) and call
  parseGinkgoFile once with the descriptor picked up front.
• Iterate over a list of required keys in
  requireClusterReportConfig instead of three identical if blocks.
• Pre-filter cluster reports once by storage key and merge
  renderClusterFailuresSection / renderMissingReportsSection into
  a single renderBulletSection helper. Output stays byte-identical;
  verified by existing messenger-report tests.
• Pass the loop credentials object as-is to postToLoopApi so the
  two call sites collapse to one-liners.

Why do we need it, and what problem does it solve?

The current E2E notification thread only lists failed test groups and
does not show why a test failed. This forces engineers to open CI
artifacts and logs even for simple failures.

The main report also counted skipped specs in the Success Rate
denominator, which made Ginkgo runs look worse than the executed-spec
result. SSH command failures contained noisy Go internals in
Failure.Message because *exec.ExitError was wrapped with %w.

When E2E fails in SynchronizedBeforeSuite, Ginkgo may skip all specs
and the JSON report may be absent or contain no regular test failures.
Previously the messenger report could only show a generic storage
failure (for example nfs with E2E TEST FAILED), without the actual
setup failure reason. The pipeline now captures Ginkgo stdout into
e2e_output_<storage>_<date>.log and the report parses suite-level
failures out of it as a fallback.

The follow-up refactors remove duplicated helpers, collapse parallel
branches, and replace inline regex/if-chain logic with declarative
data, making the report scripts easier to read and extend without
changing the rendered output.

What is the expected result?

1. Run the E2E report generation with a Ginkgo JSON report containing
   failed specs.
2. Check the main messenger report:
   • the Errors column is hidden when all error counts are zero;
   • Success Rate is calculated without skipped specs;
   • table headers include emojis.
3. Check the failed-tests thread:
   • cluster names are links;
   • the table has Tests and Reason columns;
   • Reason contains the Ginkgo Failure.Message text.
4. SSH command failures are rendered as useful text instead of dumping
   *exec.ExitError internals.
5. If Ginkgo fails in SynchronizedBeforeSuite and no JSON report is
   created, the report is still generated from the captured Ginkgo
   output and the failed-tests thread shows the suite-level failure
   reason. Suite-level failures do not change spec metrics such as
   passed, failed, errors, or total.
6. The jest suite under .github/scripts/js/e2e/report/ still passes
   (37 tests) and tmp/test-ci/report/run.sh produces the same
   markdown for the bundled fixtures.

Checklist

• The code is covered by unit tests.
• e2e tests passed.
• Documentation updated according to the changes.
• Changes were tested in the Kubernetes cluster manually.

Changelog entries

section: ci
type: fix
summary: Improve E2E messenger report failure reasons, suite setup failure reporting, and success-rate calculation.
impact_level: low