perf: improve splitting of the offset vector in GroupColumn::take_n implementations, adding a new helper fn

[mzabaluev]

Rationale for this change

The take_n implementation for ByteGroupValueBuilder uses split_vec_min_alloc to split the offset vector, then subtracts the n-th offset from the remaining offsets in place. This is slower than a single read-subtract-write pass.
The implementation in #23648 will benefit from the same optimization.

What changes are included in this PR?

Add a utility function named take_n_offsets in datafusion-common, which minimizes allocation in the way of split_vec_min_alloc, but also shifts the remaining offsets in a single pass with the copying.

Are these changes tested?

Added a micro-benchmark that shows improvement on the drain-like branch of the split, and no change on the split_off-like branch.
To keep results in context, the timed function is GroupValues::emit on a setup with two grouped columns, one of them of type Utf8, the other an integer.

Benchmark results (on a GCE instance type C4, Intel Xeon Emerald Rapids host CPU)

emit_first_small/shift_larger/i32_offsets_grp_1000
                        time:   [4.3658 µs 4.3738 µs 4.3808 µs]
                        change: [−4.6588% −3.2597% −1.9925%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 200 measurements (4.50%)
  9 (4.50%) low mild
emit_first_small/shift_midsize/i32_offsets_grp_1000
                        time:   [6.1981 µs 6.2075 µs 6.2156 µs]
                        change: [−4.5298% −3.7144% −2.8939%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 14 outliers among 200 measurements (7.00%)
  14 (7.00%) low mild
emit_first_small/shift_smaller/i32_offsets_grp_1000
                        time:   [7.6934 µs 7.7011 µs 7.7078 µs]
                        change: [−1.1020% −0.5549% −0.0432%] (p = 0.04 < 0.05)
                        Change within noise threshold.
Found 17 outliers among 200 measurements (8.50%)
  16 (8.00%) low mild
  1 (0.50%) high mild
emit_first_small/shift_larger/i64_offsets_grp_1000
                        time:   [4.8437 µs 4.8529 µs 4.8608 µs]
                        change: [−7.9716% −6.5318% −5.0240%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 32 outliers among 200 measurements (16.00%)
  3 (1.50%) low severe
  29 (14.50%) low mild
emit_first_small/shift_midsize/i64_offsets_grp_1000
                        time:   [6.5677 µs 6.5802 µs 6.5915 µs]
                        change: [−6.8527% −5.7676% −4.6713%] (p = 0.00 < 0.05)
                        Performance has improved.
emit_first_small/shift_smaller/i64_offsets_grp_1000
                        time:   [7.7735 µs 7.7847 µs 7.7950 µs]
                        change: [−2.2958% −1.7079% −1.1464%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 200 measurements (1.00%)
  2 (1.00%) high mild

emit_first_large/shift_larger/i32_offsets_grp_1000000
                        time:   [7.9151 ms 7.9374 ms 7.9629 ms]
                        change: [−5.4314% −5.0039% −4.5755%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 5 outliers among 50 measurements (10.00%)
  3 (6.00%) high mild
  2 (4.00%) high severe
emit_first_large/shift_midsize/i32_offsets_grp_1000000
                        time:   [10.184 ms 10.205 ms 10.231 ms]
                        change: [−4.5136% −4.2636% −3.9524%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 50 measurements (4.00%)
  2 (4.00%) high severe
emit_first_large/shift_smaller/i32_offsets_grp_1000000
                        time:   [7.8004 ms 7.8123 ms 7.8265 ms]
                        change: [−1.1698% −0.9187% −0.6655%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 3 outliers among 50 measurements (6.00%)
  1 (2.00%) high mild
  2 (4.00%) high severe
emit_first_large/shift_larger/i64_offsets_grp_1000000
                        time:   [8.7289 ms 8.7478 ms 8.7684 ms]
                        change: [−8.4380% −8.0534% −7.6539%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 50 measurements (8.00%)
  3 (6.00%) high mild
  1 (2.00%) high severe
emit_first_large/shift_midsize/i64_offsets_grp_1000000
                        time:   [10.850 ms 10.865 ms 10.880 ms]
                        change: [−7.2120% −6.9233% −6.6671%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 50 measurements (2.00%)
  1 (2.00%) high mild
emit_first_large/shift_smaller/i64_offsets_grp_1000000
                        time:   [8.1261 ms 8.1466 ms 8.1707 ms]
                        change: [−0.2579% +0.0712% +0.4146%] (p = 0.69 > 0.05)
                        No change in performance detected.
Found 3 outliers among 50 measurements (6.00%)
  1 (2.00%) high mild
  2 (4.00%) high severe

I set --turbo-mode=ALL_CORE_MAX in the instance parameters and pinned CPU cores to reduce variation.

Are there any user-facing changes?

utils::take_n_offsets function is added in datafusion-common.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.70%. Comparing base (3b1ce16) to head (3c2e3ba).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #23730      +/-   ##
==========================================
+ Coverage   80.69%   80.70%   +0.01%     
==========================================
  Files        1088     1089       +1     
  Lines      367766   368159     +393     
  Branches   367766   368159     +393     
==========================================
+ Hits       296781   297135     +354     
- Misses      53276    53316      +40     
+ Partials    17709    17708       -1     

☔ 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.