[C++] Optimize dictionary support in kernels/Support nulls in DictionaryUnifier

[asfimport]

ARROW-13358, ARROW-14167, ARROW-13573, etc. add support for dictionary types in various generic 'selection'-style kernels. However, the support is fairly unoptimized, repeatedly hashing and inserting the same values. Instead, we could implement more optimized support in some cases.

For example, for if_else, we could first unify both dictionaries (in such a way that the first dictionary is unchanged), then copy-with-transpose the indices of the second argument, then copy the relevant indices of the first argument (which doesn't require transposition). This would require DictionaryUnifier to support nulls in dictionaries, however.

Reporter: David Li / @lidavidm

Related issues:

• [C++] Extend type support for if_else kernel (relates to)
• [C++] Support dictionaries directly in case_when kernel (relates to)
• [C++] Support dictionaries directly in coalesce kernel (relates to)
• [R] Include unused factor levels in coalesce() and if_else() output (is related to)

_{Note: This issue was originally created as ARROW-14177. Please see the migration documentation for further details.}

[github-actions]

This issue has been marked as stale because it has had no activity in the past 365 days. Please remove the stale label or comment below, or this issue will be closed in 14 days. If this improvement is still desired but has no current owner, please add the 'Status: needs champion' label.