GH-49456: [C++] Use static key/item/value field names for map type again

[kou]

Rationale for this change

This reverts GH-49415 because the previous behavior (using static "key"/"value"/"entries" field names for map type) is better. Our specification defines the static field names.

What changes are included in this PR?

Always use "key"/"value"/"entries" for map type field names.

Are these changes tested?

Yes.

Are there any user-facing changes?

No. This just revers the unreleased change.

• GitHub Issue: [C++] Map type should always use "key"/"value"/"entries" for field names #49456

[github-actions]

⚠️ GitHub issue #49456 has been automatically assigned in GitHub to PR creator.

[kou]

@wgtmac @lidavidm @raulcd Sorry. I was wrong. GH-49416 should be reverted. GH-49416 was a backward incompatible change but it's not released yet. So this backward incompatible change will not visible for users.

[raulcd]

I've been following the conversation on the ML, this makes sense to me. Thanks @kou
I am wondering whether a comment could be added here so someone reading the code in the future doesn't have the same questions/concerns?

[kou]

Good point! I've added a comment.

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 3 benchmarking runs that have been run so far on merge-commit f2b4e34.

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.