GH-45393: [C++][Compute] Fix wrong decoding for 32-bit column in row table

[zanmato1984]

Rationale for this change

The failure reported in #45393 seems to be caused by a careless parentheses typo introduced in #45336:

arrow/cpp/src/arrow/compute/row/encode_internal.cc

Lines 281 to 282 in e32f56b

	reinterpret_cast<uint32_t*>(col_base)[i] = *reinterpret_cast<const uint32_t*>(
	rows.fixed_length_rows(start_row + i + offset_within_row));

And unfortunately our Grouper UT doesn't have cases covering this particular code path (the questioning code path is only triggered in Grouper with very restrictive conditions: the row table is fixed-length, a 32-bit key is encoded after some other keys).

What changes are included in this PR?

An UT to reproduce the issue and the fix.

Are these changes tested?

UT included.

Are there any user-facing changes?

None.

• GitHub Issue: [Benchmarking] Conbench reports regressions and errors on latest PRs #45393

[github-actions]

⚠️ GitHub issue #45393 has been automatically assigned in GitHub to PR creator.

[zanmato1984]

@ursabot please benchmark lang=R

[ursabot]

Benchmark runs are scheduled for commit 0b81f6e. Watch https://buildkite.com/apache-arrow and https://conbench.ursa.dev for updates. A comment will be posted here when the runs are complete.

[conbench-apache-arrow]

Thanks for your patience. Conbench analyzed the 3 benchmarking runs that have been run so far on PR commit 0b81f6e.

There were 2 benchmark results indicating a performance regression:

• Pull Request Run on amd64-m5-4xlarge-linux at 2025-02-10 05:25:21Z
  • file-read (R) with compression=uncompressed, dataset=nyctaxi_2010-01, file_type=feather, language=R, output_type=dataframe
  • file-read (R) with compression=uncompressed, dataset=fanniemae_2016Q4, file_type=feather, language=R, output_type=dataframe

The full Conbench report has more details.

[zanmato1984]

This should fix the regression/failure of the R benchmarks in our CI. @pitrou would you mind to take a look? Thanks.

And sorry for my careless typo in #45336 that broke CI for such a while. cc @raulcd

[zanmato1984]

The reported performance regression shouldn't be related. I'm merging this now to get our CI back to green sooner than later.

[pitrou]

Thanks for the quick fix @zanmato1984 !

[raulcd]

Thanks for the fix!

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit d3c4676.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 1 possible false positive for unstable benchmarks that are known to sometimes produce them.