[thekevinbot]

Fixes #1514

A per-table hook rerun over paths that are already known and fit one invocation runs on the calling thread again, without counting CPUs. Since the streaming dispatcher (#1502), every watch rerun spawned a dispatcher thread and a worker thread and read the cgroup CPU files before spawning the hook, which put the Python e2e watch_speed_test over its 1.1x bar. Paths still arriving from a walk keep the streaming pool.

Paired local measurement, 600 alternating samples vs inotifywait (median per-pair ratio, loaded host):

build	median ratio
main 654c4f8	1.161
before #1502 (aa0f3aa)	0.961, 1.028
this PR	1.018, 0.949

watch_speed_test on this PR, five runs: medians 0.889, 0.976, 1.062, 1.042, 0.935, all passing.

Changelog / Migrations

• Changelog fragment added under <root>/<pkg>/changelog.d/ for each changed package (or: skip-changelog trailer on a commit with reason)
• Migration fragment with the same slug added under <root>/<pkg>/migrations.d/ for each changed package (every subsection but Summary _None._ when nothing breaks)

E2E Verification

• Ran e2e suites locally for every affected SDK
• Python SDK e2e: pass (impacted on-file and watch tests)
• TypeScript SDK e2e: pass
• Rust core e2e (if applicable): N/A (cargo test --lib on_file:: pass)
• packages/python/e2e-attestations/<branch>.json receipt written if packages/python changed (just e2e-attest-python)
• packages/ts/e2e-attestations/<branch>.json receipt written if packages/ts changed (just e2e-attest-ts)
• Command(s) run: testing-conventions e2e attest 'uv run python -m pytest tests/e2e/{watch_speed,on_file_query,fanout,declared_table_name,query_contract,server_speed,jsonl_parse_speed,frontmatter_parse_speed,use_case_notebook,use_case_transcripts,sigint_interrupt}_test.py -x -q'; just e2e-attest-ts
• Result summary: Python 24 passed; TypeScript 18 passed.