ARROW-12946: [C++] String swap case kernel

[Christian8491]

This PR adds swapcase compute kernel for string. It is similar to Python str.swapcase()

[github-actions]

https://issues.apache.org/jira/browse/ARROW-12946

[edponce]

LGTM. Need to add implementation for UTF8.

[pitrou]

Thanks for doing this! Just a couple comments.

[Christian8491]

@lidavidm please enable CI

[lidavidm]

For CI, it appears that Ubuntu 20.04 provides utf8proc 2.5, but the isupper/islower functions are not provided until 2.6: https://juliastrings.github.io/utf8proc/releases/

Meanwhile, RTools35 is using utf8proc 2.4; it builds utf8proc from source due to some other issue, but I believe it's using the system headers anyways looking at the compiler command since the system headers come before the self-built utf8proc ones.

I think bumping the minimum utf8proc version and fiddling with the build flags so that the utf8proc headers take precedence will be needed.

[edponce]

An alternative solution is to use helper functions already available, refer to https://github.com/apache/arrow/blob/master/cpp/src/arrow/compute/kernels/scalar_string.cc#L1391-L1408

Related notes in Arrow w.r.t. to lower/upper casing of UTF8 can be found in:

• https://github.com/apache/arrow/blob/master/python/pyarrow/tests/test_compute.py#L668-L693
• https://github.com/apache/arrow/blob/master/cpp/src/arrow/compute/kernels/scalar_string_test.cc#L578-L601

[nealrichardson]

Meanwhile, RTools35 is using utf8proc 2.4; it builds utf8proc from source due to some other issue, but I believe it's using the system headers anyways looking at the compiler command since the system headers come before the self-built utf8proc ones.

Yeah the logs say

-- Could NOT find utf8proc (missing: utf8proc_LIB) (found suitable version "2.4.0", minimum required is "2.2.0")

which doesn't make sense.

[Christian8491]

As @edponce suggested, I replaced some utf8proc functions with helper functions. After all the CI passes. This PR is ready for review.

[pitrou]

+1, thank you @Christian8491

[edponce]

Some tests invoke the incorrect kernel (ASCII test uses utf8_capitalize and/or viceversa). These "typos" are fixed in #10869