fix(dataset): use empty/null blanks when restoring deleted rows in merge

[jan-exa]

Summary

add_blanks (used by Updater/merge_columns to re-align batches with a fragment's deleted rows) filled every deleted-row position by duplicating the batch's first value via a take selection vector (the old // For simplicity, we just use the first value for deleted rows TODO). On heavily-deleted fragments with fat variable-width values this overflows Arrow's 2 GiB 32-bit offset limit — observed in production backfill on a fragment with 191,752 of 195,879 rows deleted (LanceError(Arrow): Failed to add blanks: Offset overflow error).

Fix: interleave the real data with a 1-row type-appropriate blank instead of duplicating row 0:

// before: selection_vector.push(0)  → take() duplicates row 0's full bytes per blank
// after:  indices.push((1, 0))      → interleave([batch_column, blank_array], indices)
fn blank_array(data_type, nullable, len) -> ArrayRef
// nullable → new_null_array; non-nullable → zero/empty per type
// (primitives 0, Utf8/Binary "", recursive for Struct/List/FixedSizeList/Map/FixedSizeBinary)

Output buffers no longer scale with deleted-row count (regression test: 1 MiB first value + 100,000 blanks, asserts value_data().len() stays 1 MiB).

Write-path only and reader-equivalent: blank rows sit at deleted positions and are always masked by the deletion vector, so scans never see them — test_merge_preserves_visible_rows_with_deletions asserts visible rows are byte-identical before/after a merge over a fragment with deletions. Committed schema and on-disk format are unchanged.

Non-nullable types with no constructible blank (e.g. dictionaries) now return a descriptive NotSupported error instead of silently duplicating the first value.

Verified: cargo fmt, clippy --all --tests --benches -D warnings, updater/schema-evolution/merge test suites (65 tests) all green.

Link to Devin session: https://app.devin.ai/sessions/7cd5f358690a4d7cb4fbe13d6b5e4ad5
Requested by: @jan-exa

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring