fix: read prompt_tokens_details cache reads in TS OpenAI Chat

[opieter-aws]

Description

The TypeScript OpenAI Chat Completions path doesn't read prompt_tokens_details.cached_tokens, so prompt-cache reads were invisible in its usage

Related Issues

Part of the #3546 usage-accounting work (the py/ts Chat parity gap).

Documentation PR

N/A

Type of Change

Bug fix

Testing

Ran the TS OpenAI Chat unit suite (src/models/openai/__tests__/chat.test.ts), covering a cached-read payload surfacing cacheReadInputTokens and a zero/absent case that must not introduce the field.

Checklist

• I have read the CONTRIBUTING document
• I have reviewed and understand every line of code in this PR, including any generated by AI tools, and I can explain why it works
• My change is focused and reasonably small; I have split unrelated work into separate PRs
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[github-actions]

Assessment: Approve

Clean, well-scoped bug fix that closes the py/ts Chat parity gap for prompt-cache reads. The new guard mirrors the Responses path (mapResponsesUsage in responses-adapter.ts) exactly, so both OpenAI paths now surface cacheReadInputTokens consistently, and the field stays optional (matching the Usage type and Bedrock/Anthropic adapters).

Review Notes

• Correctness: Present-and-positive guard is right and consistent across the two OpenAI paths.
• Testing: Full-object toEqual assertions are used (good). One minor gap noted inline — the "absent" wording in the second test isn't directly exercised.
• API/Scope: Internal fix, no public API surface change; PR is small and focused.

Nice job keeping this tightly aligned with the existing Responses implementation.