fix: Improve performance of the UTF-8 string comparison logic

[dconeybe]

This PR optimizes the UTF-8 string comparison logic to improve performance and simplify the algorithm. It achieves this by directly comparing UTF-16 code units and handling surrogate pairs appropriately. The changes result in faster string comparisons without sacrificing correctness, with the added benefit of a simpler and more readable algorithm.

Highlights

• Performance Improvement: Improves the performance of UTF-8 string comparison logic, bringing it back to near its original speed before a previous fix introduced a performance degradation.
• Algorithm Simplification: The performance improvements also happily led to a simplification of the algorithm.
• UTF-16 Code Unit Comparison: The comparison logic now directly compares UTF-16 code units for efficiency, leveraging the way UTF-8 and UTF-16 represent Unicode code points.
• Surrogate Handling: The code handles surrogate pairs correctly, ensuring that strings containing surrogates are ordered appropriately relative to non-surrogate strings.

The semantics of the UTF-8 string comparison logic were originally fixed by #2275, but this fix caused a material performance degradation, which was then improved by #2299 The performance was, however, still suboptimal, and this PR further improves the speed back to close to its original speed and, serendipitously, simplifies the algorithm too.

This commit is a port of firebase/firebase-js-sdk#9143