bench: byte view rank collisions

[YUZHEthefool]

Which issue does this PR close?

None. This benchmark coverage is split out of #10605 so the optimization can be measured against the same benchmark cases on main.

Rationale for this change

The existing StringView rank benchmarks cover fixed-length inline values and mixed values. These cases add coverage for long values with a sparse external-buffer path, common prefixes, and full 16-byte prefix collisions.

What changes are included in this PR?

• Add a sparse-long string_view[0-13] case.
• Add long values with a shared 7-byte prefix, including a nullable case.
• Add long values with a shared 16-byte prefix, including a nullable case.

This PR contains benchmark coverage only. The implementation change is in #10605.

Are these changes tested?

The benchmark target was run locally with:

cargo bench -p arrow --features test_utils --bench sort_kernel -- "rank string_view"

Are there any user-facing changes?

No.

[Rich-T-kid]

thank you for splitting the PR @YUZHEthefool! LGTM