Add thorough tests for time zone transitions

[ptomato]

This adds test/thorough/timezones.mjs which tests certain invariants for time zone transitions in every supported time zone. (It doesn't use snapshots, because those might have to be updated when the IANA database updates.)

For every time zone, it takes the first five transitions after 1850 (which was about when time zones started being used) and the last two transitions before 2000 (which should cover one spring-forward and one fall-back transition in zones that use DST), and for each of those seven transitions, tests

• That each disambiguation value has the expected effect
• That .with({ offset }) works to switch between the two repeated times in a fall-back transition

This covers many different sizes of UTC offset shift, from 24 hours down to 4 seconds, with the most common being -1 hour and +1 hour.

See: #556, tc39/test262#3002

[ptomato]

@justingrant Can you think of any more time zone invariants that should be tested here?

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.83%. Comparing base (48bd5b2) to head (d66e781).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3295   +/-   ##
=======================================
  Coverage   97.83%   97.83%           
=======================================
  Files          22       22           
  Lines       10725    10725           
  Branches     1856     1856           
=======================================
  Hits        10493    10493           
  Misses        215      215           
  Partials       17       17           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Ms2ger]

This seems reasonable; I'll leave it open in case Justin wants to comment but otherwise feel free to merge

[justingrant]

Looks good. The only change I'd consider would be to apply some limit on the difference between the earlier and later instants. Other than the weird Samoa 2011 instance which could be special-cased, we should verify that the difference is not larger than N hours, which we could derive experimentally. I suspect cases > 2 hours could likely be counted on one hand.

[ptomato]

Yep, 2 hours seems about right. I'll verify these later against the implementations. There are some divergences but I'm not sure if it's because of bugs or just TZDB version discrepancies.