[C++][Compute] Normalize decimal value arguments in choose DispatchBest

[zanmato1984]

Summary

• normalize decimal value arguments in choose DispatchBest after index promotion
• require decimal choose exact matches to already share the same decimal type so binding falls back to DispatchBest when rescaling is needed
• add focused dispatch and expression-binding regression coverage for mixed-scale decimal choose calls

Root cause

choose was missing the decimal normalization step that if_else, case_when, and coalesce already apply on their best-dispatch path. Separately, the decimal varargs kernels for choose were permissive enough to satisfy exact dispatch for mixed-scale decimal arguments, which let expression binding bypass DispatchBest entirely and keep mismatched decimal value types.

What changed

After promoting the index argument to int64, ChooseFunction::DispatchBest now runs decimal normalization across the value arguments before selecting the kernel. The decimal choose kernels also now carry a match constraint that only allows exact dispatch when all decimal value arguments already have the same type, so expression binding inserts the expected casts instead of short-circuiting.

Testing

• arrow-compute-expression-test --gtest_filter='Expression.BindWithImplicitCastsForChooseOnDecimal'
• arrow-compute-scalar-if-else-test --gtest_filter='TestChooseKernel.DispatchBest'
• arrow-compute-scalar-if-else-test --gtest_filter='TestChoose.*:TestChooseKernel.*'

[github-actions]

Thanks for opening a pull request!

If this is not a minor PR. Could you open an issue for this pull request on GitHub? https://github.com/apache/arrow/issues/new/choose

Opening GitHub issues ahead of time contributes to the Openness of the Apache Arrow project.

Then could you also rename the pull request title in the following format?

GH-${GITHUB_ISSUE_ID}: [${COMPONENT}] ${SUMMARY}

or

MINOR: [${COMPONENT}] ${SUMMARY}

See also:

• Other pull requests
• Contribution Guidelines - Contributing Overview