perf(transaction): backport O(M+N) fragment matching in build_manifest onto v7.0.0

[devin-ai-integration]

Summary

Backports upstream lance-format#8210 (commit 14c5f3cc) onto the v7.0.0 tag, for a patched pylance 7.0.0 build used by the silk_node column-backfill committer.

Transaction::build_manifest for Update/Delete matched each existing dataset fragment against the transaction's fragments with a linear scan (updated_fragments.iter().find(|uf| uf.id == f.id)) → O(M×N). With ~3M dataset fragments and 8192-fragment backfill commits this dominates commit time (observed 300s → 1200s+ growth per commit). The backport replaces the scans with pre-built lookups:

let removed_ids: HashSet<u64> = removed_fragment_ids.iter().copied().collect();
let updated_by_id: HashMap<u64, &Fragment> = /* first-wins over updated_fragments */;
...
if let Some(&updated) = updated_by_id.get(&f.id) { ... }

Deviations from the upstream commit (v7 differences only, semantics identical):

• v7's overlay handling in the Update arm (updated.overlays = f.overlays.clone() + tombstone_overlay_fields under RewriteColumns) is preserved inside the map-lookup branch.
• Backported test uses v7's merged_generations field name (upstream renamed it compacted_sstables) and LanceFileVersion::V2_0 in the test helper.

Verified: cargo check -p lance clean, all 53 dataset::transaction tests pass (incl. the two new tests from lance-format#8210), rustfmt clean.

Link to Devin session: https://app.devin.ai/sessions/9589ce7177cd42afbae9db44fc352d5b
Requested by: @ryantqiu

[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