filesystem: bounded-walk 100000-entry test times out on slow Windows runners (vitest's 5 s default) #5328

src/filesystem/__tests__/bounded-walk.test.ts › search_files › "stops with an error after visiting 100000 entries" (added by #5326 for #4162) runs on vitest's default 5000 ms test timeout, but it walks MAX_WALK_ENTRIES + 1 mocked entries through the real search_files path, so its cost scales with the runner's speed.

Observed on the Windows runner (Test filesystem (Windows)):

The failure has nothing to do with the PR it failed on. The test passes about 2× under the limit on a normal run, so a slow runner pushes it over.

Expected: the test passes whatever the runner load, since its cost comes from the bound it checks and not from a bug.

Fix: give that one test an explicit timeout (for example it(name, fn, 30_000)) with a comment saying why: it walks 100,001 entries on purpose. Don't add a retry, which AGENTS.md forbids, and don't lower MAX_WALK_ENTRIES for the test, which would stop it testing the real bound. This is the same pattern as #5314, which covers the HFS+ disk-image test.