Connectathon standards follow-ups: fqm#371 watch, subset-fidelity vs official cases, DEQM profile tier

[Taleef7]

Bundle of smaller follow-ups from the HL7 Connectathon wave (PRs #293/#294; HL7 Connectathon/RESEARCH_FINDINGS_2026-07-15.md):

• Watch Date-only measurementPeriodEnd (incl. the Measure.effectivePeriod.end default) is treated as start-of-day, excluding events on the last day of the measurement period projecttacoma/fqm-execution#371 (our filed bug: date-only measurementPeriodEnd — incl. the effectivePeriod.end default — parsed as start-of-day). When fixed upstream: consider the version bump (dep-approval gate) and drop the end-of-day normalization caveat from the harness + literal-diff.ts.
• Subset-fidelity quantification: run the official MADiE test-case patients (already fetched by pnpm test:official-cases) through our authored official-subset CQL (cms122_official.cql + the CMS125 production CQL) via cql-execution and score subset agreement vs the expected MeasureReports — upgrades the 'faithful subset' claim from synthetic-verified to official-case-quantified. (Do NOT run the authored demo cms122 — it expects synthetic enrollment Conditions.)
• DEQM profile tier (post-fix(fhir): correct MeasureReport conformance semantics #294 the cheap adds are done): scoring extension, versioned resolvable measure canonical, meta.profile claim + HL7 validator run, subject-list report type, evaluatedResource; then a run against the DEQM Inferno test kit (https://flame-demo.c3ib.org/deqm-test-kit). Gate the $submit-data/$collect-data operations tier on MIE interest (Send the MIE unblock package (WebChart API contract questions) + record answers #254 question).
• Optional: re-vendor CMS122 v1.0.000 (CMS122FHIRDiabetesAssessGT9Pct, canonical + name changed upstream) — cosmetic; the 0/55 drift check showed no outcome changes.
• Send the MIE unblock package (WebChart API contract questions) + record answers #254 additions (owner): DEQM $submit-data/$collect-data server-side interest + US Quality Core 0.5.0 trajectory questions for Doug/MIE.
• Known-bad-expecteds note: our 55/55 CMS122 agreement matches the committed MADiE expecteds; the January upstream discrepancy report (Java engine numerator=1 on 6 cases) remains unreconciled with the regenerated 2026-07-14 report — revisit if upstream publishes updated comparisons.