feat: widen percentile input to use larger precision

[theirix]

Which issue does this PR close?

• Closes percentile_cont is wrong for Decimal32 with lower interpolation precision #24576

Rationale for this change

A follow-up for #24419, which introduced pure-decimal interpolation, where a smaller interpolation precision was used for Decimal32 types with Euclidean interpolation. Seems like, a smaller precision for Decimal32 is not enough, and the value was saturated to zero.

Solution - just widen i32 inputs to i64 to handle edge cases. In this case, we can use the same logical interpolation precision 10^6 everywhere.

What changes are included in this PR?

• Interpolation precision is now the only one
• scale_by_num now widens i32 inputs to i64

Are these changes tested?

• The mentioned case
• More test cases to overflow the interpolator, matched with DuckDB and DataFusion floating interpolator

Are there any user-facing changes?

[codecov-commenter]

Codecov Report

❌ Patch coverage is 79.41176% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.48%. Comparing base (32badcb) to head (bd7f54c).
⚠️ Report is 70 commits behind head on main.

Files with missing lines	Patch %	Lines
...afusion/functions-aggregate/src/percentile_cont.rs	79.41%	0 Missing and 7 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #24577      +/-   ##
==========================================
+ Coverage   81.37%   81.48%   +0.10%     
==========================================
  Files        1116     1122       +6     
  Lines      397509   404264    +6755     
  Branches   397509   404264    +6755     
==========================================
+ Hits       323461   329398    +5937     
- Misses      55110    55551     +441     
- Partials    18938    19315     +377     

☔ View full report in Codecov by Harness.
📢 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.

[theirix]

@Jefffrey could you have a look please? A follow-up for exact percentiles

[kosiew]

@theirix, thanks for working on this. The change looks good to me. Using the shared interpolation precision and widening the bounded remainder arithmetic to i64 addresses the Decimal32 issue cleanly. I just have one non-blocking suggestion around coverage for the overflow fallback path.

[kosiew]

Could we add a focused unit test that exercises the upper.sub_checked(lower) overflow fallback? For example, Decimal32 with i32::MIN and i32::MAX at a non-endpoint fraction should force this path. The new SQL tests cover the normal delta path nicely, but a test like this would also lock down the separately weighted extrema path that relies on the widened remainder arithmetic.

[theirix]

Thank you, it's a good idea - added a specialised test.

[kosiew]

🚀
@theirix
Thank you for your contribution.