chore(perplexity): Added all keys for usage metadata

[keenborder786]

• Description: Missing keys for usage metadata: According to the Perplexity API documentation, the following output token details were missing in ChatPerplexity — citation_tokens, num_search_queries, and reasoning_token — which are required for accurate cost calculation, as noted in the issue. These have now been added. Although OutputTokenDetails has restricted keys, all of them have been included for now.
• Issue: Add Perplexity cost tracking #31647

[codspeed-hq]

CodSpeed Performance Report

Merging #33480 will not alter performance

_{Comparing keenborder786:feat/preplexity_cost_tracking (4914216) with master (5acd34a)¹}

Summary

✅ 1 untouched
⏩ 33 skipped²

Footnotes

1. No successful run was found on master (f8adbbc) during the generation of this report, so 5acd34a was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

2. 33 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[ccurme]

the linter is complaining here because UsageMetadata does not support extra keys.

We should follow the existing types where possible: https://python.langchain.com/api_reference/core/messages/langchain_core.messages.ai.UsageMetadata.html

reasoning_tokens has a dedicated slot reasoning under output_token_details.

citation_tokens we can also stash in output_token_details.

num_search_queries IMO can go in response_metadata since it isn't a token count but I don't feel strongly about that.

[keenborder786]

@ccurme please check now

[ccurme]

Thanks for this, could we add a test?

[ccurme]

(nit) might make sense to only populate the keys if they are present in token_usage.

[keenborder786]

@ccurme done