test: name the run in e2e coverage tooling-error artifacts

[hivecommons-hive]

Test Improvement

main() in tests/tools/e2e-coverage-report.mjs built its failure artifact from
error.runId / error.runStatus. Only CoverageReportError carries those, and
only the three run-status/capture guards near the top of collectE2ECoverage
raise it. Every throw after the manifest has been read and validated — the
.tmp guard, the artifact-kind and run-id guards, every convertScript source-map
failure, and No src/** coverage was attributable — is a plain Error, so the
published header read:

E2E coverage: tooling-error (run null; status unknown)

even when manifest.json on disk said "runId": "run-1", "status": "passed".

.github/workflows/ci.yml cats coverage/e2e/report.txt straight into
$GITHUB_STEP_SUMMARY, and that line is the whole diagnostic. It could not
distinguish "the Playwright run failed" from "only the reporting step broke" —
two failures with opposite remedies.

This re-reads the manifest in the catch. readCoverageRun was already imported;
.catch(() => null) leaves an unreadable or absent manifest behaving exactly as
it does today.

Tests

Two cases appended to tests/e2e-coverage-report.test.mjs:

• a tooling-error artifact still names the run the manifest records — seals a
  run passed, trips the No src/** path, and asserts the JSON carries
  runId: 'run-1' / runStatus: 'passed' and the text header reads
  E2E coverage: tooling-error (run run-1; status passed). Verified
  non-vacuous: reverting only the .mjs change fails this test and no other.
• a tooling-error artifact reports no run when the manifest is unreadable —
  pins the unchanged null / null behaviour.

Scope note

Issue #1069's second completion box — a run sealed failed/cancelled reported
with that status rather than tooling-error — is not reachable and is
deliberately not implemented. collectE2ECoverage rejects every non-passed
run with a CoverageReportError before any plain-Error site runs, so the
manifest fallback can only ever observe passed, and the existing ternary on
error.runStatus already handles the sealed-failed case correctly (covered by
main labels failed runs as failed and exits nonzero). Adding the branch would
be dead code. I have said the same on the issue; nothing else is left for it to
track.

Verification

• node --test tests/e2e-coverage-report.test.mjs — 37 passed, 0 failed.
• npm run test:unit:coverage:check — exit 0 (--check-source 100,
  --check-source-regions 99).
• npx prettier --check on both changed files — exit 0.

Local revision 900592b, node v26.10.0, TZ=UTC.

Coordination

Touches only main() (tests/tools/e2e-coverage-report.mjs) and appends to
tests/e2e-coverage-report.test.mjs.

• test: make the e2e coverage artifact renderable without its build #1040 edits the same file at convertScript and the script-root resolution
  inside collectE2ECoverage; it touches no throw site and not main().
• test: fold out phantom e2e coverage regions over covered lines #1051 edits getRegionCoverage and the region union; it does not touch main().

Both append to tests/e2e-coverage-report.test.mjs at the end, so a trivial
append-order rebase is possible there; no logic overlaps.

Related Issue

Closes #1069

— hive: agent=quality backend=copilot model=claude-opus-5 copilot=1.0.88

[hivecommons-hive]

Important

Held for human review by the hive's ACMM level gate.

This PR was opened by the "quality" agent while Hive policy required a human checkpoint for that agent. Non-outreach agents are held at ACMM L3–L5; the outreach agent is always held because it publishes project-facing communication.

Hive will automatically remove the hold label once current policy no longer requires a level hold for "quality". If this is an outreach PR, a human must review it and remove the label.