test: large-map correctness tests (10k+ entries) for Len/Keys/Values/Map

[millerjp]

Summary

Len, Keys, Values, Map are benchmarked at 1000 entries but correctness-tested only at ≤ 5 entries. A future optimisation using a fixed-size sink could silently truncate and unit tests wouldn't catch it.

Scope

TestLen_LargeMap, TestKeys_LargeMap, TestValues_LargeMap, TestMap_LargeMap — each seeds 10 000 entries and asserts the return matches the seed exactly (length, set membership).

Acceptance criteria

1. Four tests, each parallel, each under 1 s.
2. -race clean.

Source: test-analyst.