fix(map-elites): isolate seeded evolutionary search randomness

[liuzhengyang699]

MapElitesOptimizer(random_seed=...) reseeds Python's global generator, and every exploration, parent-selection and mutation draw uses that same generator. Constructing another optimizer or consuming randomness in the evaluator therefore changes the candidate sequence despite the fixed seed; constructing the optimizer also changes randomness for unrelated callers.

Keep a random.Random(random_seed) instance on the optimizer and use it for all search draws. The search remains stochastic when no seed is supplied.

Validation: python -m pytest tests/src/optimizers/test_map_elites_optimizer.py tests/src/optimizers/test_optimizer_package_imports.py -q — 4 passed. Both new regressions fail on upstream: global RNG state changes, and identical seeded optimizers produce different candidate histories when the evaluator consumes random numbers. Focused Ruff correctness checks (E4,E7,E9,F) and git diff --check passed. Full Ruff still reports pre-existing typing/import-style findings in these files. No model calls.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-10T16:13:36.014328Z	13769f9	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.