Cover bounded spam actor state semantics

[alyldas]

Summary

• Add regression coverage for bounded actor-state behavior with out-of-order explicit nowMs values.
• Preserve burst timestamps across duplicate rejections so delayed unique messages keep the existing burst decision semantics.
• Keep duplicate and burst pruning on their existing decision-specific paths.
• Avoid adding schedulers, timers, background cleanup, or new public API.

Validation

• npm ci with a temporary npm cache
• npm test -- tests/index.spec.ts tests/actor-state.spec.ts with a temporary npm cache
• npm run check with a temporary npm cache

Benchmark Evidence

No package benchmark script exists for this path.

The runtime behavior is unchanged after review feedback; this branch adds regression coverage for the bounded-state compatibility case.

Compatibility Notes

• Public APIs are unchanged.
• Rejected messages still do not update actor pressure by default.
• trackRejectedAttempts keeps its existing opt-in behavior.
• Finite backward-moving nowMs compatibility is preserved for delayed unique messages after duplicate rejections.

Closes #30

BEGIN_COMMIT_OVERRIDE
fix: preserve spam burst state compatibility
END_COMMIT_OVERRIDE

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8551ee57a3

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".