Refine the code structure for percentile rank

[zanmato1984]

This refinement is based on facts that:

1. "Sort and mark duplicate" ONLY concerns:
   1. The shape of the input (array vs chunked array);
   2. The input data type and physical type;
   3. If duplicates are needed (can be simply deduced from the function options).
2. "Create rankings" ONLY concerns:
   1. The NullPartitionResult generated by "sort and mark duplicate";
   2. The ranking type (ordinal vs percentile).

Therefore we can decouple these two operations from the current Ranker, and only combine them in the concrete RankXXXMetaFunctions, which themselves can be CRTP-ed.

This refinement:

1. Maintains the functionality and performance;
2. Totally avoids virtual functions;
3. Doesn't introduce more code explosion than before;
4. Doesn't increase the complexity (IMHO it might actually reduce the complexity a bit).

[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

[pitrou]

This is neat, thank you. Can you just address the two comments I posted?

[pitrou]

Or I can just do so myself, actually.

[zanmato1984]

Thank you. Let me do this right away.

[zanmato1984]

Comments addressed. @pitrou