fix: avoid materializing constant scalar into null slots in constant page decoder

[tanish-exa]

Summary

Fixes the SEV-1271 Offset overflow error when reading Lance 2.2 files with null-padded constant pages (the poisoned Silk Node crawl_output fragments in the DLQ).

Root cause: 2.2's constant encoding stores a repeated value once, and on decode ConstantDecodeTask::materialize_values repeated the scalar for every visible slot — including null slots — via take(scalar, [0,0,...]), applying validity only afterwards. Real DLQ fragments hold one large (2–63 MB) utf8/binary value plus ~1000 nulls in the same column chunk; repeating the scalar for the null slots overflows Arrow's i32 utf8/binary offsets (>2 GiB) even though the actual data is tiny.

Fix (in constant.rs): when definition levels are present, build a 2-slot source [scalar, null] and take with per-row indices from the def levels, so null slots contribute zero bytes:

fn materialize_values(&self, def: Option<&[u16]>, num_values: u64) -> Result<ArrayRef> {
    // def-level-aware path: take from [scalar, null] instead of
    // repeating the scalar for every slot including nulls
}

The all-non-null path is unchanged (a genuinely constant 2 GiB+ column still errors, same as before).

Validation

• New regression tests test_constant_layout_null_padded_large_value_v2_2 (utf8) and test_constant_layout_null_padded_binary_v2_2 mirror the DLQ shape (32 MiB value + ~1000 nulls); both fail with Offset overflow error: 2147483648 on the old code and pass with the fix.
• cargo test -p lance-encoding: 461 passed. cargo fmt --all --check and cargo clippy --all --tests --benches -- -D warnings clean.
• Real production data: built the patched pylance wheel and read all 41 poisoned crawl_output fragments downloaded from the production DLQ — stock v9.0.1-exa.1 decodes 0/41 (all fail with Offset overflow error at constant.rs:473), patched build decodes 41/41 (full scans at batch_size=8192 plus take_rows).

Link to Devin session: https://app.devin.ai/sessions/ef2caed0e6614c15974a46d1e764e730
Requested by: @tanish-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