[mohitgurav20]

Which issue does this PR close?

• Closes Use Exact Hash-Table Allocation Accounting in GroupValuesColumn #25736

Rationale for this change

GroupValuesColumn stores its hash map as a HashTable<(u64, GroupIndexView)>,
but size() was computing the map's contribution using a hand-maintained
map_size field — incremented via insert_accounted on every insert.
That only accounts for the raw entry storage, not hashbrown's internal
control bytes or trailing layout padding.

The result: GroupValues::size() quietly underreports real memory,
especially right after a table growth. This weakens memory-pool
enforcement and spill decisions.

ArrowBytesMap already solves this correctly with HashTable::allocation_size().
This PR brings GroupValuesColumn in line with that (same approach as merged #25750 for GroupValuesRows and #25825 for GroupValuesPrimitive).

What changes are included in this PR?

Only datafusion/physical-plan/src/aggregates/group_values/multi_group_by/mod.rs is touched.

• Removed the map_size: usize field and every insert_accounted call that maintained it.
• Replaced insert_accounted with insert_unique (hashbrown's plain insert) — no behavioural change, just stops threading an accumulator through.
• GroupValuesColumn::size() now calls self.map.allocation_size() directly, matching ArrowBytesMap::size().
• Retains existing accounting for group_index_lists, emit_group_index_list_buffer, vectorized_operation_buffers, hashes_buffer, and size_of::<Self>().
• Cleaned up now-unused HashTableAllocExt imports.

What is the testing strategy for this PR?

Four focused unit tests added, one per acceptance criterion from the issue:

Test	What it verifies
map_allocation_size_included_in_size_after_growth	After 128 inserts (forces a resize), size() covers the full allocation_size()
collision_chain_size_is_separate_from_map_allocation	Forced collision: map allocation and group_index_lists are tracked independently
size_reflects_retained_capacity_after_partial_emit	EmitTo::First doesn't shrink the map allocation — retained capacity stays charged
reuse_after_full_emit_produces_correct_groups	After a full emit + reuse, group IDs restart correctly and size() stays accurate

All 122 tests in aggregates::group_values::multi_group_by pass locally.

Are there any user-facing changes?

No. This is a memory-accounting fix only. Group values, group IDs, and query
results are completely unchanged.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.82%. Comparing base (454ca08) to head (e27ed80).

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #26093      +/-   ##
==========================================
+ Coverage   82.80%   82.82%   +0.01%     
==========================================
  Files        1149     1149              
  Lines      452213   452785     +572     
  Branches   452213   452785     +572     
==========================================
+ Hits       374473   375029     +556     
- Misses      54945    54957      +12     
- Partials    22795    22799       +4     

☔ 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.