GH-45755: [C++][Python][Compute] Add winsorize function

[pitrou]

Rationale for this change

Add a "winsorize" vector function as described here:
https://en.wikipedia.org/wiki/Winsorizing

and implemented in e.g. Scipy:
https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.mstats.winsorize.html

Also make the "quantile" function supported on decimal32/decimal64.

Are these changes tested?

Yes.

Are there any user-facing changes?

No, only a new compute function.

• GitHub Issue: [C++][Compute] Add winsorize function #45755

[zanmato1984]

Some first glance questions.

Plus, no update to doc? https://github.com/apache/arrow/blob/ab4263476d9d5078cd0fa2cce6b922eb7d90c0af/docs/source/cpp/compute.rst

[pitrou]

Plus, no update to doc? https://github.com/apache/arrow/blob/ab4263476d9d5078cd0fa2cce6b922eb7d90c0af/docs/source/cpp/compute.rst

Oops, I had entirely forgotten.

[pitrou]

Ok, I've added the docs now. Do you want to take another look?

[pitrou]

@github-actions crossbow submit -g cpp

[pitrou]

@github-actions crossbow submit -g cpp

[github-actions]

Revision: be3f7aa

Submitted crossbow builds: ursacomputing/crossbow @ actions-fe0234df2b

Task	Status
example-cpp-minimal-build-static
example-cpp-minimal-build-static-system-dependency
example-cpp-tutorial
test-alpine-linux-cpp
test-build-cpp-fuzz
test-conda-cpp
test-conda-cpp-meson
test-conda-cpp-valgrind
test-cuda-cpp-ubuntu-22.04-cuda-11.7.1
test-debian-12-cpp-amd64
test-debian-12-cpp-i386
test-fedora-39-cpp
test-ubuntu-22.04-cpp
test-ubuntu-22.04-cpp-20
test-ubuntu-22.04-cpp-bundled
test-ubuntu-22.04-cpp-emscripten
test-ubuntu-22.04-cpp-no-threading
test-ubuntu-24.04-cpp
test-ubuntu-24.04-cpp-bundled-offline
test-ubuntu-24.04-cpp-gcc-13-bundled
test-ubuntu-24.04-cpp-gcc-14
test-ubuntu-24.04-cpp-minimal-with-formats
test-ubuntu-24.04-cpp-thread-sanitizer

[zanmato1984]

LGTM with two nits.

[icexelloss]

@pitrou Do you think There is any benefit to resolve and use the quantile kernel here directly as supposed to use CallFunction?

I suppose it is easier this way (using CallFunction), but I wonder, in general, when writing a kernel that uses other kernel/functions, whether it is better to use CallFunction or resolve it kernel and use kernel->Exec

[pitrou]

Resolving the kernel could perhaps save some nanoseconds, but I'm not sure that's significant compared to the other costs.

[icexelloss]

Left a question. Otherwise LGTM.