test: cover CaseStudies' undated-study comparator arms end to end

[hivecommons-hive]

Test Improvement

sortByPublishedAtDesc compares two case studies through a fallback on each
side, (b.publishedAt || '').localeCompare(a.publishedAt || '')
(src/components/CaseStudies/index.js line 25). All 213 studies in
data/case-studies.json carry a publishedAt, so neither arm can run, and the
empty arm of the date cell beside it (line 204) cannot render.

A missing date is a property of one record rather than a cleared
document-level field, so this uses the additive overlay mechanism
(tests/e2e/fixtures/data/case-studies.json, see
tests/tools/e2e-data-fixtures.cjs) rather than the variant build: the real
studies keep rendering exactly as they did and the undated rows render
alongside them, so the arm is added rather than swapped. npm run build:production and the gating end-to-end job never load the overlay.

The overlay appends two undated studies. A lone undated record only ever
lands on one side of a comparison against the dated ones, leaving the fallback
on the other side unreached — measured at 94.83% with line 25 still listed as
uncovered, because that line carries two regions. With two records the pair is
also compared against each other and both fallbacks run.

tests/e2e/case-studies.spec.js gains one case asserting that each undated row
renders with an empty date cell and its neighbouring cells intact, that the
undated studies occupy the final rows (which is what shows the comparator ran
its fallbacks rather than dropping the rows), and that the other rows still
show a formatted date.

Evidence

Measured at main 900592b by building both coverage sites
(npm run build:e2e:coverage) and running the instrumented suite locally
(npm run test:e2e:coverage with E2E_COVERAGE_DIR/E2E_COVERAGE_RUN_ID
set), then sealing and rendering the report:

	regions	covered	CaseStudies	src files
baseline (298 passed)	468	379	92.98%, uncovered [25, 46, 64]	80.98%
this PR (299 passed)	467	380	96.55%, uncovered [46, 64]	81.37%

The baseline render matches job End-to-end coverage of run
37164361552
attempt 1 to within one region (CI 468 / 379 / 80.98%), and CaseStudies is
identical in both.

The result is unchanged under the region fold proposed in #1051
(87.73% -> 88.37% src files, CaseStudies 96.55% either way), so this does
not depend on which of the two lands first and does not touch its files.

The CI gates pass as configured in .github/workflows/ci.yml:
report:e2e:coverage --check-source 100 --check-source-regions 80 --require-source-files exits 0, as do npm run test:unit:coverage:check,
npm run test:unit and npm run check:format.

Not in this PR

The two remaining regions, lines 46 (!data.generatedAt) and 64
(data.caseStudies || []), are document-level fields that an additive overlay
cannot reach. They are out of scope here and called out on the issue.

Related Issue

Closes #1064

---

Filed by quality agent (hold-gated mode). Human review required.

— 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.