test(vector): assert routing invariant, not exact HNSW recall (#773)

[mosuka]

Summary

vector_field_routing_test.rs asserted the presence of every expected doc (e.g. ids.contains(&1) && ids.contains(&2)) on a randomized HNSW graph. That is an exact-recall assertion, which is environment-sensitively flaky: the Test laurus (ubuntu-latest, x86_64-unknown-linux-gnu, stable) CI job intermittently returned only the nearest doc ({1}) and failed at vector_field_routing_test.rs:138, while the aarch64 and Windows jobs of the same run passed. It is not reproducible locally (75/75 passes here), consistent with HNSW build-RNG/connectivity non-determinism — the anti-pattern lessons.md (2026-05-29) already documents.

All five tests in the file shared this pattern, so all were latent flakes; only test_query_vector_fields_routes happened to fire.

Change

These are routing tests. The deterministic guarantee is that a query routed to a field returns only that field's docs (no cross-field leakage) — not that approximate search recalls every one. So:

• Single-field routing tests now assert the result is non-empty and a subset of the targeted field's docs (drops the flaky per-doc contains, keeps the no-leak guarantee).
• test_no_fields_searches_all asserts both fields are represented, anchored on the docs identical to the query (doc 1 in img_vec, doc 3 in txt_vec, similarity 1.0, reliably returned) instead of requiring all four docs.

The observed failure case {1} now passes deterministically ({1} ⊆ {1,2} and non-empty), so this resolves the flake without weakening the routing guarantee.

Verification

• cargo fmt --check — clean
• cargo clippy -p laurus --test vector_field_routing_test -- -D warnings — clean
• Ran the test file 40× in a loop — 40/40 pass (was intermittently failing on CI).

Closes #773