Add disableRowMaterialization option to skip per-cell decode

[mani-mathur-arch]

Summary

Adds an opt-in connection option, disableRowMaterialization (default false), that lets a caller drain a result set for its row count and fetch/parse cost without paying the per-cell row-materialization overhead.

When enabled, ArrowResultConverter still fetches, decompresses and IPC-parses every Arrow batch, but skips table.toArray() + getRows() (the recursive convertArrowTypes → convertThriftTypes pass) and returns null row placeholders instead. Only the row count is meaningful on this path — the returned objects carry no cell values.

Motivation

The Node driver materializes every cell into a JS object on every fetch path, which dominates wall-clock on large result sets and makes throughput benchmarks measure decode cost rather than fetch cost. This flag gives benchmarks (and any count-only consumer) a way to isolate fetch/parse from per-row decode. It does not change any default behavior.

Changes

• disableRowMaterialization?: boolean added to ConnectionOptions (IDBSQLClient.ts) and ClientConfig (IClientContext.ts), defaulted false in DBSQLClient.getDefaultConfig and copied from connect() options — same plumbing as preserveBigNumericPrecision.
• ArrowResultConverter.fetchNext returns new Array(rowCount).fill(null) when the flag is set (filled, not sparse, so fetchAll's .flat() preserves the count).
• Wired through both backends: Thrift (ARROW_BASED_SET + URL_BASED_SET/CloudFetch) and kernel.

What it does not do

The flag skips per-cell materialization, not the Arrow IPC batch parse (RecordBatchReader.from / new Table), which still runs. So it brings fetch to an "Arrow batches parsed, rows not materialized" level, identical on both the Thrift and kernel backends.

How is this tested?

• New unit test in tests/unit/result/ArrowResultConverter.test.ts — returns null placeholders with exact counts when disableRowMaterialization is set — mirrors the existing multi-batch "should respect row count in batch" case and asserts each fetch returns the exact row count with every element null.
• Verified locally: full build clean, eslint clean on changed files, tsc --noEmit adds no new errors (the pre-existing examples/tokenFederation module-resolution errors are present on main too), and 155 unit tests pass across the touched suites (ArrowResultConverter, ResultSlicer, DBSQLClient, DBSQLOperation).

[github-actions]

Thanks for your contribution! To satisfy the DCO policy in our contributing guide every commit message must include a sign-off message. One or more of your commits is missing this message. You can reword previous commit messages with an interactive rebase (git rebase -i main).

[peco-review-bot]

✅ No issues identified by the review bot.

[peco-review-bot]

✅ No issues identified by the review bot.

[msrathore-db]

Add support for kernel mode