GH-15281: [C++] Replace bytes_view alias with span

[bkietz]

Rationale for this change

basic_string_view was never really intended to support types other than chars; the template which the STL intends for use in this situation is std::span

What changes are included in this PR?

Since std::span is added in c++20, a minimal polyfill is added named util::span.

• doesn't include support for static extents
• some accessors like front() are not present in util::span

User facing changes:

• The broken alias bytes_view and its header are removed. I don't expect this to trouble anybody much since it isn't used in much of our API, and if it's really necessary it will be trivial to add it back with using bytes_view = span<uint8_t const>; or so

• Closes: [C++] basic_string_view<long> will be invalid in future libc++ #15281

[pitrou]

@benibus Do you want to review this?

[pitrou]

Should also close #29678.

[bkietz]

To workaround #36379 for this PR I could add an overload of FormatMatcherDescription which forwards to the v1.11 one when necessary but won't be called otherwise. However that seems pretty hacky. Please advise

[pitrou]

Let's not overengineer it. You just want to check the pointer/length of a span, so you can write a trivial helper function AssertSpansEqual (or AssertSpansIdentical if you want to stress that it's not a logical comparison of the underlying values).

[felipecrv]

@bkietz you remove an existing public header. Shouldn't that be mentioned in the PR description as an user-facing change?

[benibus]

Thanks, this will be useful! In general, LGTM - just one nit.

[pitrou]

Should this be R&&?

[bkietz]

Alright, I'll add a more tests to check construction cases against std::span

[pitrou]

@github-actions crossbow submit -g cpp

[pitrou]

@github-actions crossbow submit -g cpp

[pitrou]

@github-actions crossbow submit -g cpp

[github-actions]

Revision: 2e3a9eb

Submitted crossbow builds: ursacomputing/crossbow @ actions-972a30de43

Task	Status
test-alpine-linux-cpp
test-build-cpp-fuzz
test-conda-cpp
test-conda-cpp-valgrind
test-cuda-cpp
test-debian-11-cpp-amd64
test-debian-11-cpp-i386
test-fedora-35-cpp
test-ubuntu-20.04-cpp
test-ubuntu-20.04-cpp-bundled
test-ubuntu-20.04-cpp-minimal-with-formats
test-ubuntu-20.04-cpp-thread-sanitizer
test-ubuntu-22.04-cpp
test-ubuntu-22.04-cpp-20

[pitrou]

Made a small improvement to the tests, I'll merge now. Thank you @bkietz !

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 6 benchmarking runs that have been run so far on merge-commit 366dbe1.

There were no benchmark performance regressions. 🎉

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