Stage 2.7 review feedback: Be more specific about safe date ranges

[sffc]

Several of our calendars currently use modern methods, but not long ago, they used other methods on the ground.

For example, the modern Thai Buddhist calendar was apparently not adopted until 1940¹, and the modern Chinese calendar methods were not adopted until approximately 1953, when the Purple Mountain Observatory published their almanac.²

For answering the question "how far back we should support", here is my favorite rule of thumb: anyone currently alive should be able to express their birthday. The oldest person currently alive was born in 1909. (Note: The oldest Japanese person currently alive was born in 1911, the last full year of the Meiji era.) To give a nice round number, we can say that dates starting in 1900 ought to be accurate according to the ground truth.

For projecting into the future, I don't exactly know, but let's say 100-200 years for now. The Umm al-Qura tables appear to go through about 2173 CE. I don't know how far out official Chinese tables go. I expect that this window could change over time, but perhaps only with a change to our specification.

Note: This question is about defining the behavior of the calendars within a certain window of time. It does not take a position on the behavior outside of that safe range. See tc39/proposal-temporal#2869.

Footnotes

1. https://unicode-org.atlassian.net/browse/ICU-23162 ↩

2. https://ytliu0.github.io/ChineseCalendar/rules.html ↩

[sffc]

I largely implemented this in #65. I don't think fixing the rest of this is blocking Stage 2.7. At least now we have a certain range of dates for which we can write tests.