Gregorian (and related) epochs are off by one from the civil Julian Day Number

[hosni]

DateObject#toJulianDay() is toDays() + calendar.epoch. For Gregorian dates the result is one less than the standard civil Julian Day Number (ICU / astronomical midnight-based civil JDN).

Date	Current toJulianDay	Civil JDN
Gregorian 0001-01-01	1721425	1721426
Gregorian 2024-03-20	2460389	2460390

Jalali is a special case: with the current Birashk leapsLength it also lands on 2460389 for 1403-01-01, so Jalali↔Gregorian conversions stay consistent while both are wrong relative to civil JDN.

After fixing Jalali leapsLength to match isLeap (see #21 and #22), Jalali 1403-01-01 becomes 2460390 (civil-correct) while Gregorian stays at 2460389, so the two calendars diverge by one day until this fix lands.

Reproduction

const gregorian = require("./calendars/cjs/gregorian");

function jd(cal, y, m, d) {
  return cal.getAllDays({ year: y, month: { index: m - 1 }, day: d }) + cal.epoch;
}

console.log(jd(gregorian, 2024, 3, 20)); // 2460389 (civil: 2460390)

Suggested fix

1. Add 1 to these epochs: gregorian, julian, indian, arabic
   (leave jalali / persian at 1948319 once leapsLength matches isLeap).
2. Update the weekday formula in index.js so Wednesday stays Wednesday when JDNs move to the civil scale:

// before
let index = (this.toJulianDay() + 3) % 7;
// after
let index = (this.toJulianDay() + 2) % 7;

Impact

Breaking change for callers who stored raw Gregorian/Julian/Indian/Arabic toJulianDay() values.
Improves ICU / civil-JDN interoperability.