[C++][Compute] Tighten if_else exact dispatch for parameterized value types

[zanmato1984]

Summary

• tighten if_else exact dispatch so the value operands must fully match as DataTypes before an exact kernel can be selected
• keep DispatchBest responsible for decimal normalization / cast insertion instead of letting broad type_id matches short-circuit that path
• add targeted dispatch and expression-binding regressions for decimal casts plus exact-dispatch rejects for timestamp timezone, fixed-size binary, nested, and dictionary mismatches

Root cause

if_else registered several exact kernels with signatures that only matched the value operands by type_id (for example decimals, fixed-size binary, nested, dictionary, and timestamp unit matchers). That let exact dispatch succeed for parameterized value types that were not actually type-equal, which could bypass the normalization and cast-insertion logic that lives in DispatchBest and expression binding.

What changed

• add a shared MatchConstraint for if_else kernels that requires then and else to have the same full DataType during exact dispatch
• apply that constraint to the registered primitive, binary, fixed-width, and nested if_else kernels so exact dispatch no longer accepts broader parameterized matches
• add decimal-focused DispatchBest and expression binding tests to confirm the cast path still runs when exact dispatch is rejected

Validation

• ./build/debug/arrow-compute-scalar-if-else-test --gtest_filter=TestIfElseKernel.IfElseDispatchBest:TestIfElseKernel.IfElseDispatchExact
• ./build/debug/arrow-compute-expression-test --gtest_filter=Expression.BindWithImplicitCastsForIfElseOnDecimal

[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