Polyfill: Replace unreachable overshoot block with assertion in calendarToIsoDate

[jessealama]

The overshoot block in calendarToIsoDate's binary search appears unreachable. The closest candidate is Coptic/Ethiopic month 13 (only 5–6 days), where the 8-day step could theoretically skip the month entirely — but even there, I haven't been able to construct an example that triggers it. The refinement step keeps estimates close enough that calculateSameMonthResult always handles them first.

Replace it with an assertNotReached so we get a clear error if this analysis is ever wrong.

Closes #3217

[jessealama]

This PR is similar to #3291 in that it originates in a more aggressive trimming of dead code (#3289)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.83%. Comparing base (13f6fab) to head (faf750a).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3292      +/-   ##
==========================================
+ Coverage   97.64%   97.83%   +0.19%     
==========================================
  Files          22       22              
  Lines       10746    10725      -21     
  Branches     1855     1856       +1     
==========================================
  Hits        10493    10493              
+ Misses        235      215      -20     
+ Partials       18       17       -1     

☔ 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.

[ptomato]

Thanks for doing the extra research to try to find a case that would reach this code. I was a bit uneasy removing this because it may have been put there for a reason that's lost to the mists of time (this generic calendar implementation was done in a hurry)

On the other hand, since it corresponds to the implementation-defined part of the spec text, there isn't that much potential downside to removing it, as you point out.

I think the assertNotReached can be simplified into an assert so we don't have to mess around with c8 magic comments, but I'll fix that up before merging.

Thanks!

[ptomato]

We should fix up the comment as well, since it's no longer a binary search.