GH-50994: [C++][Compute] Implement casting from ListView to List with zero-copy fast-path

[Jay846]

Rationale for this change

This PR implements missing type-casting compute kernels to convert ListViewType and LargeListViewType arrays to standard ListType and LargeListType arrays. This directly addresses legacy cast removal requirements under Arrow's compute framework.

What changes are included in this PR?

To maximize performance and optimize memory layouts, a dual-execution path was implemented in the CastListView execution functor inside scalar_cast_nested.cc:

• Contiguous Zero-Copy Fast-Path: Triggered when the input list-view elements are contiguous and adjacent (i.e. offsets[i] + sizes[i] == offsets[i+1]). It avoids copying the child values array entirely, allocating the new output offset buffer, shifting offsets relative to the start, and slicing the child array directly to preserve zero-copy pointer semantics.
• Non-Contiguous Slower Path: Triggered for gapped, overlapping, or out-of-order layouts. It dynamically tracks target offsets, builds integer index mappings using an Int64Builder, and invokes Arrow's internal take compute kernel to reconstruct a new contiguous child values array.
• Kernel Registration: Formally wired into Arrow's casting system for standard and large variants of both nested types.

Are these changes tested?

Yes, added comprehensive unit test suites in scalar_cast_test.cc passing 109 out of 109 cases in the scalar cast suite. Tests explicitly cover:

• Contiguous Arrays: Validated zero-copy memory behavior programmatically by ensuring the source and destination child buffers share the exact same physical memory address.
• Gapped & Overlapping Arrays: Verified correct offset tracking and reconstruction when offsets intersect or contain unused slots.
• Null Propagation: Checked that null validity maps propagate correctly at both the parent and child levels.
• Generic/Nested Types: Validated casting on non-primitive data types (e.g. utf8 strings) and recursive child type promotions (e.g., ListView<int16> to List<int32>).
• Boundary Slices: Verified that taking slices of non-contiguous arrays computes correct offsets and avoids out-of-bound errors.

Are there any user-facing changes?

No public API contracts were broken. This adds declarative casting support natively to the existing internal compute framework.

• GitHub Issue: GH-50994: [C++][Compute] Implement casting from ListView to List with zero-copy fast-path #50976
• GitHub Issue: [C++][Compute] Implement casting from ListView/LargeListView to List/LargeList #50994

[github-actions]

Thanks for opening a pull request!

This pull request has been automatically converted to a draft because its title doesn't match Arrow's required format.

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}

After updating the title, you can mark the pull request as ready for review.

See also:

• Other pull requests
• Contribution Guidelines - Contributing Overview

[Reranko05]

Could you use the Arrow PR title template:-

GH-<Issue Number>: [<Component>] <Title>

[github-actions]

⚠️ GitHub issue #50976 has been automatically assigned in GitHub to PR creator.

[HuaHuaY]

You should use issue id but not the PR's id. If there isn't an existed issue, you can create a new one.

[github-actions]

⚠️ GitHub issue #50994 has been automatically assigned in GitHub to PR creator.