Optimized add_all and upsert_all operations in ordered_map module

[PersonaNormale]

Performance Optimization for OrderedMap Bulk Operations

Problem

The current implementation of add_all and upsert_all in ordered_map performs individual binary searches for each element in an unsorted array.

Solution

This PR replaces vector::zip approach with a quick sort-based algorithm that:

• Pre-sorts key-value pairs together in a single pass
• Tracks insertion positions to minimize redundant binary searches
• Properly handles error cases for mismatched arrays

Performance Improvement

Before:

• Vector Element Insertion = O(entries)
• Find Index = O(log(entries))
  And we were forced to do this for vec_elements times.
• Leading to: O(vec_elements * log(entries) * entries)

After:
Now, that we have modified the binary search for reusing last index because we have warranty of an ordered vector:

• Sorting = O(vec_elements log(vec_elements))
• Insertion = O(entries * log(entries - last_index))
  That can lead to an O(vec_elements log(vec_elements)) if we have a lot of elements
  Or O(entries * log(entries - last_index_found)) if entries is greater or vec_elements are already sorted/psuedo-sorted.
• In total = O(vec_elements log(vec_elements) + entries * log(entries - last_index_found))

Implementation Details

• Added quick_sort_with_companion_array implementation using Hoare's partition algorithm
• Added is_ordered helper function to detect already-sorted arrays, worst case for quicksort
• Added error constant EARRAYS_DIFFERENT_LENGTH (error code 5)

Testing

Added test coverage for:

• Edge cases (empty arrays, single elements, already sorted data)
• Error handling for mismatched arrays and invalid inputs
• Various input data patterns and sizes
• Validation of sorting correctness

All existing tests continue to pass with no regressions.

[trunk-io]

⏱️ 6s total CI duration on this PR

Job	Cumulative Duration	Recent Runs
permission-check	3s	🟥
permission-check	3s	🟥

_{settings ⋅ feedback ⋅ docs ⋅ learn more about trunk.io}