Relax error checking requiring exclusive end in NudgeToCalendarUnit

[arshaw]

Fix for #2919

Looks like the sanity checking was a little too aggressive. Was asserting a range with an exclusive end. Should be inclusive end instead. To see what I mean, I'll breaking down the operations from #2919 ...

// original recreation
const dur = Temporal.Duration.from({ months: 11 });
const relativeTo = Temporal.PlainDate.from("2023-05-31");
dur.round({ relativeTo, smallestUnit: "months", roundingMode: "ceil" });

// granular operations
const durBalanced = relativeTo.until(relativeTo.add(dur), { largestUnit: 'months' }); // P10M30D
const months0 = 10 // because we trunc P10M30D down to 10 months
const months1 = 11 // because 10 plus roundingInc (1) is 11 months
const windowPd0 = relativeTo.add({ months: months0 }) // 2024-03-31
const windowPd1 = relativeTo.add({ months: months1 }) // 2024-04-30
const destPd = relativeTo.add(dur) // 2024-04-30

/*
You then must determine how far `destPd` is between `windowPd0` <-> `windowPd1`
As you can see, `destPd` === `windowPd1`, which should be allowed
*/

[codecov]

Codecov Report

Attention: Patch coverage is 66.66667% with 1 line in your changes missing coverage. Please review.

Project coverage is 96.26%. Comparing base (82ad54f) to head (c0bd886).
Report is 54 commits behind head on main.

Files	Patch %	Lines
polyfill/lib/ecmascript.mjs	66.66%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2924      +/-   ##
==========================================
- Coverage   96.51%   96.26%   -0.26%     
==========================================
  Files          23       23              
  Lines       12432    12073     -359     
  Branches     2258     2206      -52     
==========================================
- Hits        11999    11622     -377     
- Misses        374      387      +13     
- Partials       59       64       +5     

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

[arshaw]

Updated per #2919 (comment)
CC @ptomato and @anba

[ptomato]

Thanks @arshaw, I'll add the corresponding spec text to this branch. @anba, your input will be appreciated before we merge this.

[ptomato]

I wrote a test case that passes with this branch and fails on main: tc39/test262#4196

[anba]

I can't test if the proposed changes will always compute the expected results until I've updated SpiderMonkey's Temporal implementation to use the current DifferenceISODate. This won't happen before finishing the changes for #2925, so it'll probably take some time.

[ptomato]

Thanks for taking a look. I've changed the assertion back to an exception for the time being, though, as you say, it may need to change back to an assertion after #2925 lands. We should land this only after #2925, I think.

[ptomato]

I'm going to roll this one into #2925 because test262 tests have already been merged for it. That also allows dropping the "Editorial: Downgrade assertion back to exception" commit, because without custom calendars that invariant can no longer be broken.

[ptomato]

Merged as part of #2925.