perf: improve ArrayUnique lookup performance

[itamar-gerasy]

Replaces the filter and indexOf implementation in arrayUnique with Set-based duplicate detection.

For arrays containing distinct values, the current implementation has O(n²) worst-case time complexity. The new implementation uses O(n) expected time while preserving identifier-based comparison and reference-based object comparison.

This also fixes the handling of NaN by using JavaScript's SameValueZero equality semantics:

• a single NaN, or one NaN among otherwise distinct values, is accepted;
• multiple NaN values are rejected as duplicates;
• sparse array holes are treated as undefined, following normal Set iteration semantics.

Tests were added for object reference comparison, NaN, sparse arrays, and identifier evaluation.

Checklist

• the pull request title describes what this PR does (not a vague title like Update index.md)
• the pull request targets the default branch of the repository (develop)
• the code follows the established code style of the repository
  • npm run prettier:check passes
  • npm run lint:check passes
• tests are added for the changes I made (if any source code was modified)
• documentation added or updated (not applicable; the public API and documented behavior are unchanged)
• I have run the project locally and verified that there are no errors

Validation completed successfully:

• 677 functional tests
• ES2015 build
• ESM5 build
• CommonJS build
• UMD build
• Type declarations build

Fixes

Fixes #2690