HijriCalendar.toHijri() previously used new Date(y, m, d) (local-time
constructor). Under hijri-core's new UTC-day contract the engine reads
the UTC calendar day, so on east-of-UTC hosts (e.g. UTC+5) the local
midnight falls on the previous UTC date, producing a one-day-off Hijri
result.
Fix: use Date.UTC(y, m, d) so PlainDate calendar fields land in the
Date's UTC components, matching what hijri-core reads.
Lock-step dependency: this fix requires the unreleased hijri-core change
on fix/utc-day-boundary (commit 3419378). Both packages will be released
together per ADR-013.
Tests: round-trip regression added (2025-03-01 = 1 Ramadan 1446 AH);
all 37 ESM + 9 CJS tests pass at TZ=UTC, TZ=America/New_York,
TZ=Pacific/Auckland.
Fix documentation style (no em dashes). Update hijri-core devDep
from file: path to ^1.0.0. Add wiki-sync workflow permissions.
Tighten CalendarEngine interface usage; correct month range docs
to 1318-1500 AH throughout.
