[thekevinbot]

Fixes #1474

SELECT path FROM './**/*.md' now runs at or under fd.

• Streaming rows. The scan hands rows to the cursor in chunks of 4096 while the walk is still running (row_stream.rs). Before, the whole walk finished before the row loop began.
• Static text. Path, basename and ext are stored NUL-terminated and handed to SQLite as SQLITE_STATIC text. That saves one copy per cell.

Measured with paired A/B on a 1.36M-file tree (median of 11 pairs, host load ~9):

Build	Ratio vs fd
Base	1.27
This PR	0.90

On 4 CPUs (taskset, 2 cores with SMT), this PR runs at 0.80x the base build.

An earlier commit capped the walk at physical cores. It is reverted here: on a 4-vCPU runner the cap halved the walk workers, and the walk ran 1.8x slower than the previous release. Locally, on 4 CPUs, 2 workers took 1.94x the base build's time and 4 workers took 0.85x. With streaming rows, the cap is not needed to beat fd.

Two other levers were measured and dropped because they did not pay off:

• Fewer get_ref FFI calls: ~12ms, and not reachable through rusqlite.
• Pre-sizing the output buffer: no gain.

SELECT path output matches fd byte for byte (md5 over the sorted paths).

set_terminated_text gets a line-scoped mutation exemption in testing-conventions.toml. It writes through a raw sqlite3_context, and only a real Connection supplies one. tests/vtab.rs (stat_columns_carry_real_values) kills both of its mutants.

Changelog / Migrations

• Changelog fragment added under packages/rust/changelog.d/ for each changed package
• Migration fragment with the same slug added under packages/rust/migrations.d/ for each changed package (every subsection but Summary _None._)

E2E Verification

• Ran e2e suites locally for every affected SDK
• Python SDK e2e: pass (61 passed, every e2e test except watch_speed_test)
• TypeScript SDK e2e: pass (18 passed)
• Rust core e2e: pass (cargo test --workspace --features cli)
• packages/python/e2e-attestations/<branch>.json receipt written
• packages/ts/e2e-attestations/<branch>.json receipt written (just e2e-attest-ts)
• Command(s) run: testing-conventions e2e attest 'uv run python -m pytest tests/e2e/ --ignore=tests/e2e/watch_speed_test.py -q' (in packages/python, after maturin develop --release and cargo build --release -p dirsql --features cli), just e2e-attest-ts, cargo test --workspace --features cli
• Result summary: watch_speed_test was left out on purpose. It measures write-to-event latency on a one-file tree through dirsql server, a path this change does not touch, and it is currently red on main for an unrelated reason.