[thekevinbot]

Refs #1534

Takes single-threaded work off the critical path of a config-table build whose per-table hook starts during the walk:

• stream_table_rows shapes each chunk of hook rows on its own thread (db::shape_json_rows, split out of Db::shape_rows). Shaping one chunk now overlaps inserting the previous one.
• The grouping thread also settles the early tables' file lists (split_streamed). The batches loop no longer walks those files after the inserts.
• Prepare no longer copies each scan result (Found) into a ScannedFile. The grouping thread makes the copy (Scanned::into_files).
• Db::get_table_columns quotes the table name. The new code reads columns before any rows arrive, which broke a table named order that matched no files.

Avoids tree_walk.rs, scanner.rs and listing.rs (#1521, #1528).

Measurements

Corpus: 131072 notes. Paired, alternating runs; medians. The machine was shared, load 18-30.

	ratio
this PR vs main, release binary, A/B (31, 61, 61 pairs)	0.957, 0.969, 0.980
on top of #1521, A/B (41 pairs)	0.952
vs fd+xargs -P (fastest native), main / this PR, 31 pairs	2.34 / 2.20
frontmatter_parse_speed_test, fastest (15 pairs)	1.70

The goal (at most 1.0x the fastest native) is not met, so the test stays on the find+awk bar (it passes at 0.18). Timing a build of this PR on top of #1521 (131k files, ~125ms total):

• walk: ~33-48ms
• last hook invocation ends: ~77-93ms (about native wall time)
• inserts end: ~20ms after that
• query (GROUP BY over 131k rows): ~21-25ms

What is left needs the walk files (#1521's sys CPU cut, then more) plus a way around the ~20ms SQLite GROUP BY, which the natives do not pay.

Changelog / Migrations

• Changelog fragment added under packages/rust/changelog.d/
• Migration fragment with the same slug added under packages/rust/migrations.d/

E2E Verification

• Ran e2e suites locally for every affected SDK
• Python SDK e2e: pass (impacted subset, 22 passed)
• TypeScript SDK e2e: pass (18 passed)
• Rust core e2e: pass (cargo test --workspace --features cli, 2068 passed)
• packages/python/e2e-attestations/bot-1534-fm-native-1016.json receipt written
• packages/ts/e2e-attestations/bot-1534-fm-native-1016.json receipt written
• Command(s) run: testing-conventions e2e attest 'uv run python -m pytest tests/e2e/frontmatter_parse_speed_test.py tests/e2e/jsonl_parse_speed_test.py tests/e2e/on_file_query_test.py tests/e2e/fanout_test.py tests/e2e/declared_table_name_test.py tests/e2e/query_contract_test.py tests/e2e/use_case_notebook_test.py tests/e2e/use_case_transcripts_test.py tests/e2e/sigint_interrupt_test.py -q -s'; just e2e-attest-ts; cargo test --workspace --features cli
• Result summary: all pass. Ratios against the fastest native: frontmatter 1.70, jsonl 1.36, notebook 0.90, transcripts 0.62.