Normative: Limit Duration units to prevent arbitrary-precision integer arithmetic, and prevent loops

[ptomato]

This is the integer math refactor that I have mentioned in the last couple of Temporal presentations.

It is stacked on top of #2519 and #2609; only the last 7 commits belong to this change. The GitHub UI doesn't let me show a comparison with HEAD~7 but here is a view that includes the editorial commits from #2609 but without the large PR #2519: https://github.com/tc39/proposal-temporal/compare/user-code-calls...duration-normalize-96?expand=1

Best reviewed commit by commit. There is more information in each individual commit message.

Test262 tests are in https://github.com/ptomato/test262/commits/duration-normalize but I still need to review them for complete coverage.

See: #2195, #1700.

This is now part 3 of 3 of this change.
Part 1: #2722
Part 2: #2727

[codecov]

Codecov Report

Attention: 8 lines in your changes are missing coverage. Please review.

Comparison is base (7bd3589) 96.59% compared to head (891769a) 96.53%.

❗ Current head 891769a differs from pull request most recent head a57b62a. Consider uploading reports for the commit a57b62a to get more accurate results

Files	Patch %	Lines
polyfill/lib/ecmascript.mjs	81.39%	7 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2612      +/-   ##
==========================================
- Coverage   96.59%   96.53%   -0.06%     
==========================================
  Files          23       23              
  Lines       12218    12243      +25     
  Branches     2263     2268       +5     
==========================================
+ Hits        11802    11819      +17     
- Misses        356      364       +8     
  Partials       60       60              

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

[ptomato]

Draft until presented at TC39.

cc: @syg, @anba, @rkirsling

[Mrashes]

Few comments after looking over the PR - new around here but wanted to contribute!

[ptomato]

This is now split into 2 parts. Part 1 is in #2722 and the test262 tests are in tc39/test262#3957.

[ptomato]

The last thing I needed to investigate on this one was some odd behaviour that @cjtenny and @anba discovered when the UTC offset shift is > 24 hours in a custom time zone. (e.g, your UTC offset changes from -13:00 to +13:00, meaning a UTC offset shift of +26 hours.) A shift this large doesn't exist in any actual time zones on earth. (±24:00 is the maximum, on the rare occasions when a region decides to jump to the other side of the International Date Line.)

So what I've done here is to add checks forbidding such transitions. The only place such a transition occurs is in contrived custom time zones, where we also have the precedent of forbidding similar shenanigans (see #2725 for example.) If such a transition ever occurs in future in the real world (quite unlikely), we'd lift those restrictions and also consider running this loop an extra time.

I think all concerns have now been addressed on part 3/3 of this PR and it's ready to be merged once it gets a positive review. Tests are in tc39/test262#3999, which also need a positive review.